diff --git a/master/annotated.html b/master/annotated.html index c1251c57db..2af7c85d4a 100644 --- a/master/annotated.html +++ b/master/annotated.html @@ -243,12 +243,11 @@  CPanel  CPanel< axis, const T, D, StoreTransposed::No >  CPanel< axis, const T, D, StoreTransposed::Yes > - CRetiledMatrix - CSubTileSpecContains the information to create a subtile - CTile - CTile< const T, D > - CSubMatrixView - CSubPanelView + CSubTileSpecContains the information to create a subtile + CTile + CTile< const T, D > + CSubMatrixView + CSubPanelView  ►Nmemory  CMemoryChunkThe class MemoryChunk represents a layer of abstraction over the underlying device memory  CMemoryView diff --git a/master/auxiliary_2norm_2mc_8h_source.html b/master/auxiliary_2norm_2mc_8h_source.html index 79617de5e1..1f1d7e2f98 100644 --- a/master/auxiliary_2norm_2mc_8h_source.html +++ b/master/auxiliary_2norm_2mc_8h_source.html @@ -219,7 +219,7 @@
148 }
149 }
dlaf::internal::Policy
Definition: policy.h:24
-
dlaf::matrix::Matrix
Definition: matrix.h:55
+
dlaf::matrix::Matrix
Definition: matrix.h:56
dlaf::common::make_data
auto make_data(Data &&data) noexcept
Definition: data.h:78
distribution.h
tile.h
diff --git a/master/band__to__tridiag_8h.html b/master/band__to__tridiag_8h.html index 1af7e6f608..0d57ca0f0e 100644 --- a/master/band__to__tridiag_8h.html +++ b/master/band__to__tridiag_8h.html @@ -151,7 +151,9 @@

23 namespace dlaf {
24 namespace eigensolver {
25 
-
73 template <Backend B, Device D, class T>
-
74 TridiagResult<T, Device::CPU> bandToTridiag(blas::Uplo uplo, SizeType band_size,
-
75  Matrix<const T, D>& mat_a) {
-
76  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
-
77  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
-
78  DLAF_ASSERT(mat_a.blockSize().rows() % band_size == 0, mat_a.blockSize().rows(), band_size);
-
79  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
-
80  DLAF_ASSERT(band_size >= 2, band_size);
-
81 
-
82  switch (uplo) {
-
83  case blas::Uplo::Lower:
-
84  return internal::BandToTridiag<B, D, T>::call_L(band_size, mat_a);
-
85  break;
-
86  case blas::Uplo::Upper:
-
87  DLAF_UNIMPLEMENTED(uplo);
-
88  break;
-
89  case blas::Uplo::General:
-
90  DLAF_UNIMPLEMENTED(uplo);
-
91  break;
-
92  }
-
93 
-
94  return DLAF_UNREACHABLE(TridiagResult<T, Device::CPU>);
-
95 }
-
96 
-
144 template <Backend backend, Device device, class T>
-
145 TridiagResult<T, Device::CPU> bandToTridiag(comm::CommunicatorGrid grid, blas::Uplo uplo,
-
146  SizeType band_size, Matrix<const T, device>& mat_a) {
-
147  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
-
148  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
-
149  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
-
150  DLAF_ASSERT(band_size >= 2, band_size);
-
151 
-
152  // If the grid contains only one rank force local implementation.
-
153  if (grid.size() == comm::Size2D(1, 1))
-
154  return bandToTridiag<backend, device, T>(uplo, band_size, mat_a);
+
74 template <Backend B, Device D, class T>
+
75 TridiagResult<T, Device::CPU> bandToTridiag(blas::Uplo uplo, SizeType band_size,
+
76  Matrix<const T, D>& mat_a) {
+
77  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
+
78  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
+
79  DLAF_ASSERT(mat_a.blockSize().rows() % band_size == 0, mat_a.blockSize().rows(), band_size);
+
80  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
+
81  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
+
82  DLAF_ASSERT(band_size >= 2, band_size);
+
83 
+
84  switch (uplo) {
+
85  case blas::Uplo::Lower:
+
86  return internal::BandToTridiag<B, D, T>::call_L(band_size, mat_a);
+
87  break;
+
88  case blas::Uplo::Upper:
+
89  DLAF_UNIMPLEMENTED(uplo);
+
90  break;
+
91  case blas::Uplo::General:
+
92  DLAF_UNIMPLEMENTED(uplo);
+
93  break;
+
94  }
+
95 
+
96  return DLAF_UNREACHABLE(TridiagResult<T, Device::CPU>);
+
97 }
+
98 
+
147 template <Backend backend, Device device, class T>
+
148 TridiagResult<T, Device::CPU> bandToTridiag(comm::CommunicatorGrid grid, blas::Uplo uplo,
+
149  SizeType band_size, Matrix<const T, device>& mat_a) {
+
150  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
+
151  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
+
152  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
+
153  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
+
154  DLAF_ASSERT(band_size >= 2, band_size);
155 
-
156  switch (uplo) {
-
157  case blas::Uplo::Lower:
-
158  return internal::BandToTridiag<backend, device, T>::call_L(grid, band_size, mat_a);
-
159  break;
-
160  case blas::Uplo::Upper:
-
161  DLAF_UNIMPLEMENTED(uplo);
-
162  break;
-
163  case blas::Uplo::General:
-
164  DLAF_UNIMPLEMENTED(uplo);
-
165  break;
-
166  }
-
167 
-
168  return DLAF_UNREACHABLE(TridiagResult<T, Device::CPU>);
-
169 }
-
170 
-
171 }
-
172 }
-
dlaf::eigensolver::bandToTridiag
TridiagResult< T, Device::CPU > bandToTridiag(blas::Uplo uplo, SizeType band_size, Matrix< const T, D > &mat_a)
Definition: band_to_tridiag.h:74
+
156  // If the grid contains only one rank force local implementation.
+
157  if (grid.size() == comm::Size2D(1, 1))
+
158  return bandToTridiag<backend, device, T>(uplo, band_size, mat_a);
+
159 
+
160  switch (uplo) {
+
161  case blas::Uplo::Lower:
+
162  return internal::BandToTridiag<backend, device, T>::call_L(grid, band_size, mat_a);
+
163  break;
+
164  case blas::Uplo::Upper:
+
165  DLAF_UNIMPLEMENTED(uplo);
+
166  break;
+
167  case blas::Uplo::General:
+
168  DLAF_UNIMPLEMENTED(uplo);
+
169  break;
+
170  }
+
171 
+
172  return DLAF_UNREACHABLE(TridiagResult<T, Device::CPU>);
+
173 }
+
174 
+
175 }
+
176 }
+
dlaf::eigensolver::bandToTridiag
TridiagResult< T, Device::CPU > bandToTridiag(blas::Uplo uplo, SizeType band_size, Matrix< const T, D > &mat_a)
Definition: band_to_tridiag.h:75
dlaf::comm::CommunicatorGrid
Definition: communicator_grid.h:42
dlaf::comm::CommunicatorGrid::size
Size2D size() const noexcept
Return the number of rows in the grid.
Definition: communicator_grid.h:68
diff --git a/master/bt__band__to__tridiag_8h_source.html b/master/bt__band__to__tridiag_8h_source.html index b35d41624f..91c970c24f 100644 --- a/master/bt__band__to__tridiag_8h_source.html +++ b/master/bt__band__to__tridiag_8h_source.html @@ -121,43 +121,48 @@
51 // @pre band_size is a divisor of mat_hh.blockSize().cols()
52 // @pre mat_e is not distributed
53 // @pre mat_hh is not distributed
-
54 template <Backend B, Device D, class T>
-
55 void backTransformationBandToTridiag(const SizeType band_size, matrix::Matrix<T, D>& mat_e,
-
56  matrix::Matrix<const T, Device::CPU>& mat_hh) {
-
57  DLAF_ASSERT(matrix::local_matrix(mat_e), mat_e);
-
58  DLAF_ASSERT(matrix::local_matrix(mat_hh), mat_hh);
-
59 
-
60  DLAF_ASSERT(matrix::square_size(mat_hh), mat_hh);
-
61  DLAF_ASSERT(matrix::square_blocksize(mat_hh), mat_hh);
-
62 
-
63  DLAF_ASSERT(mat_hh.size().rows() == mat_e.size().rows(), mat_hh, mat_e);
-
64  DLAF_ASSERT(mat_hh.blockSize().rows() == mat_e.blockSize().rows(), mat_hh, mat_e);
-
65 
-
66  DLAF_ASSERT(band_size >= 2, band_size);
-
67  DLAF_ASSERT(mat_hh.blockSize().rows() % band_size == 0, mat_hh.blockSize(), band_size);
-
68 
-
69  internal::BackTransformationT2B<B, D, T>::call(band_size, mat_e, mat_hh);
-
70 }
-
71 
-
72 template <Backend B, Device D, class T>
-
73 void backTransformationBandToTridiag(comm::CommunicatorGrid grid, const SizeType band_size,
-
74  matrix::Matrix<T, D>& mat_e,
-
75  matrix::Matrix<const T, Device::CPU>& mat_hh) {
-
76  DLAF_ASSERT(matrix::equal_process_grid(mat_e, grid), mat_e, grid);
-
77  DLAF_ASSERT(matrix::equal_process_grid(mat_hh, grid), mat_hh, grid);
-
78 
-
79  DLAF_ASSERT(matrix::square_size(mat_hh), mat_hh);
-
80  DLAF_ASSERT(matrix::square_blocksize(mat_hh), mat_hh);
-
81 
-
82  DLAF_ASSERT(mat_hh.size().rows() == mat_e.size().rows(), mat_hh, mat_e);
-
83  DLAF_ASSERT(mat_hh.blockSize().rows() == mat_e.blockSize().rows(), mat_hh, mat_e);
-
84 
-
85  DLAF_ASSERT(band_size >= 2, band_size);
-
86  DLAF_ASSERT(mat_hh.blockSize().rows() % band_size == 0, mat_hh.blockSize(), band_size);
-
87 
-
88  internal::BackTransformationT2B<B, D, T>::call(grid, band_size, mat_e, mat_hh);
-
89 }
-
90 }
+
54 // @pre mat_e has equal tile and block sizes
+
55 // @pre mat_hh has equal tile and block sizes
+
56 template <Backend B, Device D, class T>
+
57 void backTransformationBandToTridiag(const SizeType band_size, matrix::Matrix<T, D>& mat_e,
+
58  matrix::Matrix<const T, Device::CPU>& mat_hh) {
+
59  DLAF_ASSERT(matrix::local_matrix(mat_e), mat_e);
+
60  DLAF_ASSERT(matrix::local_matrix(mat_hh), mat_hh);
+
61 
+
62  DLAF_ASSERT(matrix::square_size(mat_hh), mat_hh);
+
63  DLAF_ASSERT(matrix::square_blocksize(mat_hh), mat_hh);
+
64 
+
65  DLAF_ASSERT(mat_hh.size().rows() == mat_e.size().rows(), mat_hh, mat_e);
+
66  DLAF_ASSERT(mat_hh.blockSize().rows() == mat_e.blockSize().rows(), mat_hh, mat_e);
+
67 
+
68  DLAF_ASSERT(matrix::single_tile_per_block(mat_e), mat_e);
+
69  DLAF_ASSERT(matrix::single_tile_per_block(mat_hh), mat_hh);
+
70 
+
71  DLAF_ASSERT(band_size >= 2, band_size);
+
72  DLAF_ASSERT(mat_hh.blockSize().rows() % band_size == 0, mat_hh.blockSize(), band_size);
+
73 
+
74  internal::BackTransformationT2B<B, D, T>::call(band_size, mat_e, mat_hh);
+
75 }
+
76 
+
77 template <Backend B, Device D, class T>
+
78 void backTransformationBandToTridiag(comm::CommunicatorGrid grid, const SizeType band_size,
+
79  matrix::Matrix<T, D>& mat_e,
+
80  matrix::Matrix<const T, Device::CPU>& mat_hh) {
+
81  DLAF_ASSERT(matrix::equal_process_grid(mat_e, grid), mat_e, grid);
+
82  DLAF_ASSERT(matrix::equal_process_grid(mat_hh, grid), mat_hh, grid);
+
83 
+
84  DLAF_ASSERT(matrix::square_size(mat_hh), mat_hh);
+
85  DLAF_ASSERT(matrix::square_blocksize(mat_hh), mat_hh);
+
86 
+
87  DLAF_ASSERT(mat_hh.size().rows() == mat_e.size().rows(), mat_hh, mat_e);
+
88  DLAF_ASSERT(mat_hh.blockSize().rows() == mat_e.blockSize().rows(), mat_hh, mat_e);
+
89 
+
90  DLAF_ASSERT(band_size >= 2, band_size);
+
91  DLAF_ASSERT(mat_hh.blockSize().rows() % band_size == 0, mat_hh.blockSize(), band_size);
+
92 
+
93  internal::BackTransformationT2B<B, D, T>::call(grid, band_size, mat_e, mat_hh);
+
94 }
+
95 }
util_matrix.h
diff --git a/master/bt__reduction__to__band_8h.html b/master/bt__reduction__to__band_8h.html index a8331c8c60..c3dcbe081c 100644 --- a/master/bt__reduction__to__band_8h.html +++ b/master/bt__reduction__to__band_8h.html @@ -147,7 +147,11 @@

Precondition
mat_c is distributed,
-mat_v is distributed according to grid.
+mat_v is distributed according to grid, +
+mat_c has equal tile and block sizes,
+
+mat_v has equal tile and block sizes.
@@ -202,7 +206,11 @@

Precondition
mat_c is not distributed,
-mat_v is not distributed.
+mat_v is not distributed, +
+mat_c has equal tile and block sizes,
+
+mat_v has equal tile and block sizes.
diff --git a/master/bt__reduction__to__band_8h_source.html b/master/bt__reduction__to__band_8h_source.html index b5bca2d98a..23e2b5e049 100644 --- a/master/bt__reduction__to__band_8h_source.html +++ b/master/bt__reduction__to__band_8h_source.html @@ -92,52 +92,56 @@
22 namespace dlaf {
23 namespace eigensolver {
24 
-
38 template <Backend backend, Device device, class T>
-
39 void backTransformationReductionToBand(const SizeType b, Matrix<T, device>& mat_c,
-
40  Matrix<const T, device>& mat_v,
-
41  Matrix<const T, Device::CPU>& mat_taus) {
-
42  DLAF_ASSERT(matrix::local_matrix(mat_c), mat_c);
-
43  DLAF_ASSERT(matrix::local_matrix(mat_v), mat_v);
-
44  DLAF_ASSERT(square_size(mat_v), mat_v);
-
45  DLAF_ASSERT(square_blocksize(mat_v), mat_v);
-
46  DLAF_ASSERT(mat_c.size().rows() == mat_v.size().rows(), mat_c, mat_v);
-
47  DLAF_ASSERT(mat_c.blockSize().rows() == mat_v.blockSize().rows(), mat_c, mat_v);
-
48 
-
49  [[maybe_unused]] auto nr_reflectors_blocks = [&b, &mat_v]() {
-
50  const SizeType m = mat_v.size().rows();
-
51  const SizeType mb = mat_v.blockSize().rows();
-
52  return std::max<SizeType>(0, util::ceilDiv(m - b - 1, mb));
-
53  };
-
54  DLAF_ASSERT(mat_taus.nrTiles().rows() == nr_reflectors_blocks(), mat_taus.size().rows(), mat_v, b);
-
55 
-
56  internal::BackTransformationReductionToBand<backend, device, T>::call(b, mat_c, mat_v, mat_taus);
-
57 }
-
58 
-
72 template <Backend backend, Device device, class T>
-
73 void backTransformationReductionToBand(comm::CommunicatorGrid grid, const SizeType b,
-
74  Matrix<T, device>& mat_c, Matrix<const T, device>& mat_v,
-
75  Matrix<const T, Device::CPU>& mat_taus) {
-
76  DLAF_ASSERT(matrix::equal_process_grid(mat_c, grid), mat_c, grid);
-
77  DLAF_ASSERT(matrix::equal_process_grid(mat_v, grid), mat_v, grid);
-
78  DLAF_ASSERT(square_size(mat_v), mat_v);
-
79  DLAF_ASSERT(square_blocksize(mat_v), mat_v);
-
80  DLAF_ASSERT(mat_c.size().rows() == mat_v.size().rows(), mat_c, mat_v);
-
81  DLAF_ASSERT(mat_c.blockSize().rows() == mat_v.blockSize().rows(), mat_c, mat_v);
-
82 
-
83  [[maybe_unused]] auto nr_reflectors_blocks = [&b, &mat_v]() {
-
84  const SizeType m = mat_v.size().rows();
-
85  const SizeType mb = mat_v.blockSize().rows();
-
86  return mat_v.distribution().template nextLocalTileFromGlobalTile<Coord::Col>(
-
87  std::max<SizeType>(0, util::ceilDiv(m - b - 1, mb)));
-
88  };
-
89  DLAF_ASSERT(mat_taus.distribution().localNrTiles().rows() == nr_reflectors_blocks(), mat_taus, mat_v,
-
90  b);
-
91 
-
92  internal::BackTransformationReductionToBand<backend, device, T>::call(grid, b, mat_c, mat_v, mat_taus);
-
93 }
-
94 }
-
95 }
-
dlaf::eigensolver::backTransformationReductionToBand
void backTransformationReductionToBand(const SizeType b, Matrix< T, device > &mat_c, Matrix< const T, device > &mat_v, Matrix< const T, Device::CPU > &mat_taus)
Definition: bt_reduction_to_band.h:39
+
40 template <Backend backend, Device device, class T>
+
41 void backTransformationReductionToBand(const SizeType b, Matrix<T, device>& mat_c,
+
42  Matrix<const T, device>& mat_v,
+
43  Matrix<const T, Device::CPU>& mat_taus) {
+
44  DLAF_ASSERT(matrix::local_matrix(mat_c), mat_c);
+
45  DLAF_ASSERT(matrix::local_matrix(mat_v), mat_v);
+
46  DLAF_ASSERT(square_size(mat_v), mat_v);
+
47  DLAF_ASSERT(square_blocksize(mat_v), mat_v);
+
48  DLAF_ASSERT(mat_c.size().rows() == mat_v.size().rows(), mat_c, mat_v);
+
49  DLAF_ASSERT(mat_c.blockSize().rows() == mat_v.blockSize().rows(), mat_c, mat_v);
+
50  DLAF_ASSERT(single_tile_per_block(mat_c), mat_c);
+
51  DLAF_ASSERT(single_tile_per_block(mat_v), mat_v);
+
52 
+
53  [[maybe_unused]] auto nr_reflectors_blocks = [&b, &mat_v]() {
+
54  const SizeType m = mat_v.size().rows();
+
55  const SizeType mb = mat_v.blockSize().rows();
+
56  return std::max<SizeType>(0, util::ceilDiv(m - b - 1, mb));
+
57  };
+
58  DLAF_ASSERT(mat_taus.nrTiles().rows() == nr_reflectors_blocks(), mat_taus.size().rows(), mat_v, b);
+
59 
+
60  internal::BackTransformationReductionToBand<backend, device, T>::call(b, mat_c, mat_v, mat_taus);
+
61 }
+
62 
+
78 template <Backend backend, Device device, class T>
+
79 void backTransformationReductionToBand(comm::CommunicatorGrid grid, const SizeType b,
+
80  Matrix<T, device>& mat_c, Matrix<const T, device>& mat_v,
+
81  Matrix<const T, Device::CPU>& mat_taus) {
+
82  DLAF_ASSERT(matrix::equal_process_grid(mat_c, grid), mat_c, grid);
+
83  DLAF_ASSERT(matrix::equal_process_grid(mat_v, grid), mat_v, grid);
+
84  DLAF_ASSERT(square_size(mat_v), mat_v);
+
85  DLAF_ASSERT(square_blocksize(mat_v), mat_v);
+
86  DLAF_ASSERT(mat_c.size().rows() == mat_v.size().rows(), mat_c, mat_v);
+
87  DLAF_ASSERT(mat_c.blockSize().rows() == mat_v.blockSize().rows(), mat_c, mat_v);
+
88  DLAF_ASSERT(single_tile_per_block(mat_c), mat_c);
+
89  DLAF_ASSERT(single_tile_per_block(mat_v), mat_v);
+
90 
+
91  [[maybe_unused]] auto nr_reflectors_blocks = [&b, &mat_v]() {
+
92  const SizeType m = mat_v.size().rows();
+
93  const SizeType mb = mat_v.blockSize().rows();
+
94  return mat_v.distribution().template nextLocalTileFromGlobalTile<Coord::Col>(
+
95  std::max<SizeType>(0, util::ceilDiv(m - b - 1, mb)));
+
96  };
+
97  DLAF_ASSERT(mat_taus.distribution().localNrTiles().rows() == nr_reflectors_blocks(), mat_taus, mat_v,
+
98  b);
+
99 
+
100  internal::BackTransformationReductionToBand<backend, device, T>::call(grid, b, mat_c, mat_v, mat_taus);
+
101 }
+
102 }
+
103 }
+
dlaf::eigensolver::backTransformationReductionToBand
void backTransformationReductionToBand(const SizeType b, Matrix< T, device > &mat_c, Matrix< const T, device > &mat_v, Matrix< const T, Device::CPU > &mat_taus)
Definition: bt_reduction_to_band.h:41
dlaf::comm::CommunicatorGrid
Definition: communicator_grid.h:42
communicator_grid.h
dlaf::eigensolver::eigensolver
void eigensolver(blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)
Definition: eigensolver.h:39
diff --git a/master/cholesky_8h.html b/master/cholesky_8h.html index 95ccbbe5aa..762c1b99b7 100644 --- a/master/cholesky_8h.html +++ b/master/cholesky_8h.html @@ -129,6 +129,8 @@

22 namespace dlaf {
23 namespace factorization {
24 
-
38 template <Backend backend, Device device, class T>
-
39 void cholesky(blas::Uplo uplo, Matrix<T, device>& mat_a) {
-
40  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
-
41  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
-
42  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
-
43 
-
44  if (uplo == blas::Uplo::Lower)
-
45  internal::Cholesky<backend, device, T>::call_L(mat_a);
-
46  else
-
47  internal::Cholesky<backend, device, T>::call_U(mat_a);
-
48 }
-
49 
-
64 template <Backend backend, Device device, class T>
-
65 void cholesky(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, device>& mat_a) {
-
66  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
-
67  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
-
68  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
-
69 
-
70  // Method only for Lower triangular matrix
-
71  if (uplo == blas::Uplo::Lower)
-
72  internal::Cholesky<backend, device, T>::call_L(grid, mat_a);
-
73  else
-
74  internal::Cholesky<backend, device, T>::call_U(grid, mat_a);
-
75 }
-
76 
-
77 }
-
78 }
-
dlaf::factorization::cholesky
void cholesky(blas::Uplo uplo, Matrix< T, device > &mat_a)
Definition: cholesky.h:39
+
39 template <Backend backend, Device device, class T>
+
40 void cholesky(blas::Uplo uplo, Matrix<T, device>& mat_a) {
+
41  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
+
42  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
+
43  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
+
44  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
+
45 
+
46  if (uplo == blas::Uplo::Lower)
+
47  internal::Cholesky<backend, device, T>::call_L(mat_a);
+
48  else
+
49  internal::Cholesky<backend, device, T>::call_U(mat_a);
+
50 }
+
51 
+
67 template <Backend backend, Device device, class T>
+
68 void cholesky(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, device>& mat_a) {
+
69  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
+
70  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
+
71  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
+
72  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
+
73 
+
74  // Method only for Lower triangular matrix
+
75  if (uplo == blas::Uplo::Lower)
+
76  internal::Cholesky<backend, device, T>::call_L(grid, mat_a);
+
77  else
+
78  internal::Cholesky<backend, device, T>::call_U(grid, mat_a);
+
79 }
+
80 
+
81 }
+
82 }
+
dlaf::factorization::cholesky
void cholesky(blas::Uplo uplo, Matrix< T, device > &mat_a)
Definition: cholesky.h:40
dlaf::comm::CommunicatorGrid
Definition: communicator_grid.h:42
communicator_grid.h
diff --git a/master/classdlaf_1_1matrix_1_1_matrix-members.html b/master/classdlaf_1_1matrix_1_1_matrix-members.html index e060e8fb26..e0a473e6ad 100644 --- a/master/classdlaf_1_1matrix_1_1_matrix-members.html +++ b/master/classdlaf_1_1matrix_1_1_matrix-members.html @@ -89,9 +89,10 @@ readwrite(const LocalTileIndex &index) noexceptdlaf::matrix::Matrix< T, D >inline readwrite(const GlobalTileIndex &index) noexceptdlaf::matrix::Matrix< T, D >inline ReadWriteSenderType typedef (defined in dlaf::matrix::Matrix< T, D >)dlaf::matrix::Matrix< T, D > - subPipeline() (defined in dlaf::matrix::Matrix< T, D >)dlaf::matrix::Matrix< T, D >inline - TileDataType typedef (defined in dlaf::matrix::Matrix< T, D >)dlaf::matrix::Matrix< T, D > - TileType typedef (defined in dlaf::matrix::Matrix< T, D >)dlaf::matrix::Matrix< T, D > + retiledSubPipeline(const LocalTileSize &tiles_per_block)dlaf::matrix::Matrix< T, D >inline + subPipeline()dlaf::matrix::Matrix< T, D >inline + TileDataType typedef (defined in dlaf::matrix::Matrix< T, D >)dlaf::matrix::Matrix< T, D > + TileType typedef (defined in dlaf::matrix::Matrix< T, D >)dlaf::matrix::Matrix< T, D >
diff --git a/master/classdlaf_1_1matrix_1_1_matrix.html b/master/classdlaf_1_1matrix_1_1_matrix.html index e537d00dbc..d21a88e613 100644 --- a/master/classdlaf_1_1matrix_1_1_matrix.html +++ b/master/classdlaf_1_1matrix_1_1_matrix.html @@ -128,9 +128,10 @@   ReadWriteSenderType readwrite (const GlobalTileIndex &index) noexcept   - -Matrix subPipeline () +Matrix subPipeline ()   +Matrix retiledSubPipeline (const LocalTileSize &tiles_per_block) +  @@ -430,6 +431,68 @@

Tile with local index index.

Precondition
index.isIn(distribution().localNrTiles()).
+ + + +

◆ retiledSubPipeline()

+ +
+
+
+template<class T , Device D>
+

Public Attributes

+ + + + +
+ + + + + + + + +
Matrix dlaf::matrix::Matrix< T, D >::retiledSubPipeline (const LocalTileSize & tiles_per_block)
+ 		+inline
+
+

Create a sub-pipelined, retiled matrix which can be accessed thread-safely with respect to the original matrix

+

All accesses to the sub-pipelined matrix are sequenced after previous accesses and before later accesses to the original matrix, independently of when tiles are accessed in the sub-pipelined matrix.

+
Precondition
blockSize() is divisible by tiles_per_block
+
+blockSize() == baseTileSize()
+ +
+ + +

◆ subPipeline()

+ +
+
+
+template<class T , Device D>
+ + + + + +
+ + + + + + + +
Matrix dlaf::matrix::Matrix< T, D >::subPipeline ()
+ 		+inline
+
+

Create a sub-pipelined matrix which can be accessed thread-safely with respect to the original matrix

+

All accesses to the sub-pipelined matrix are sequenced after previous accesses and before later accesses to the original matrix, independently of when tiles are accessed in the sub-pipelined matrix.

+
The documentation for this class was generated from the following file:
    diff --git a/master/classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4-members.html b/master/classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4-members.html index 319dcdde6a..fbc1c50649 100644 --- a/master/classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4-members.html +++ b/master/classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4-members.html @@ -78,6 +78,8 @@ ConstTileType typedef (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > device (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >static distribution() const noexceptdlaf::matrix::internal::MatrixBaseinline + done(const LocalTileIndex &index) noexceptdlaf::matrix::Matrix< const T, D >inline + done(const GlobalTileIndex &index) noexceptdlaf::matrix::Matrix< const T, D >inline ElementType typedef (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > Matrix(const LayoutInfo &layout, ElementType *ptr) (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > Matrix(const LayoutInfo &layout, const ElementType *ptr) (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >inline @@ -87,32 +89,35 @@ Matrix(Matrix &&rhs)=default (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > Matrix(Distribution distribution) (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >inlineprotected Matrix(Matrix &mat, const SubPipelineTag) (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected - Matrix< ElementType, D > (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > - MatrixBase(Distribution distribution) (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline - MatrixBase(const Distribution &distribution, const LocalTileSize &tiles_per_block) (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline - MatrixBase(const MatrixBase &rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBase - MatrixBase(MatrixBase &&rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseprotected - nrTiles() const noexceptdlaf::matrix::internal::MatrixBaseinline - operator=(const Matrix &rhs)=delete (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > - operator=(Matrix &&rhs)=default (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > - operator=(const MatrixBase &rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBase - operator=(MatrixBase &&rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseprotected - rankGlobalTile(const GlobalTileIndex &global_tile) const noexceptdlaf::matrix::internal::MatrixBaseinline - rankIndex() const noexceptdlaf::matrix::internal::MatrixBaseinline - read(const LocalTileIndex &index) noexceptdlaf::matrix::Matrix< const T, D >inline - read(const GlobalTileIndex &index)dlaf::matrix::Matrix< const T, D >inline - ReadOnlySenderType typedef (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > - setUpSubPipelines(Matrix< const T, D > &) noexcept (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected - setUpTiles(const memory::MemoryView< ElementType, D > &mem, const LayoutInfo &layout) noexcept (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected - size() const noexceptdlaf::matrix::internal::MatrixBaseinline - sourceRankIndex() const noexcept (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline - subPipelineConst() (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >inline - tile_managers_ (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected - TileDataType typedef (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > - tileLinearIndex(const LocalTileIndex &index) const noexceptdlaf::matrix::internal::MatrixBaseinlineprotected - tileSize(const GlobalTileIndex &index) const noexceptdlaf::matrix::internal::MatrixBaseinline - TileType typedef (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > - waitLocalTiles() noexceptdlaf::matrix::Matrix< const T, D > + Matrix(Matrix &mat, const LocalTileSize &tiles_per_block) (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected + Matrix< ElementType, D > (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > + MatrixBase(Distribution distribution) (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline + MatrixBase(const Distribution &distribution, const LocalTileSize &tiles_per_block) (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline + MatrixBase(const MatrixBase &rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBase + MatrixBase(MatrixBase &&rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseprotected + nrTiles() const noexceptdlaf::matrix::internal::MatrixBaseinline + operator=(const Matrix &rhs)=delete (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > + operator=(Matrix &&rhs)=default (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > + operator=(const MatrixBase &rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBase + operator=(MatrixBase &&rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseprotected + rankGlobalTile(const GlobalTileIndex &global_tile) const noexceptdlaf::matrix::internal::MatrixBaseinline + rankIndex() const noexceptdlaf::matrix::internal::MatrixBaseinline + read(const LocalTileIndex &index) noexceptdlaf::matrix::Matrix< const T, D >inline + read(const GlobalTileIndex &index)dlaf::matrix::Matrix< const T, D >inline + ReadOnlySenderType typedef (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > + retiledSubPipelineConst(const LocalTileSize &tiles_per_block)dlaf::matrix::Matrix< const T, D >inline + setUpRetiledSubPipelines(Matrix< const T, D > &, const LocalTileSize &tiles_per_block) noexcept (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected + setUpSubPipelines(Matrix< const T, D > &) noexcept (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected + setUpTiles(const memory::MemoryView< ElementType, D > &mem, const LayoutInfo &layout) noexcept (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected + size() const noexceptdlaf::matrix::internal::MatrixBaseinline + sourceRankIndex() const noexcept (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline + subPipelineConst()dlaf::matrix::Matrix< const T, D >inline + tile_managers_ (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D >protected + TileDataType typedef (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > + tileLinearIndex(const LocalTileIndex &index) const noexceptdlaf::matrix::internal::MatrixBaseinlineprotected + tileSize(const GlobalTileIndex &index) const noexceptdlaf::matrix::internal::MatrixBaseinline + TileType typedef (defined in dlaf::matrix::Matrix< const T, D >)dlaf::matrix::Matrix< const T, D > + waitLocalTiles() noexceptdlaf::matrix::Matrix< const T, D >
    diff --git a/master/classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html b/master/classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html index c30a5680a1..fb7388cfb7 100644 --- a/master/classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html +++ b/master/classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html @@ -143,9 +143,14 @@   void waitLocalTiles () noexcept   - -Matrix subPipelineConst () +Matrix subPipelineConst ()   +Matrix retiledSubPipelineConst (const LocalTileSize &tiles_per_block) +  +void done (const LocalTileIndex &index) noexcept +  +void done (const GlobalTileIndex &index) noexcept +  - Public Member Functions inherited from dlaf::matrix::internal::MatrixBase  MatrixBase (Distribution distribution) @@ -217,12 +222,18 @@  Matrix (Matrix &mat, const SubPipelineTag)   + + Matrix (Matrix &mat, const LocalTileSize &tiles_per_block) +  void setUpTiles (const memory::MemoryView< ElementType, D > &mem, const LayoutInfo &layout) noexcept   void setUpSubPipelines (Matrix< const T, D > &) noexcept   + +void setUpRetiledSubPipelines (Matrix< const T, D > &, const LocalTileSize &tiles_per_block) noexcept +  - Protected Member Functions inherited from dlaf::matrix::internal::MatrixBase  MatrixBase (MatrixBase &&rhs)=default @@ -240,6 +251,66 @@  

    Member Function Documentation

    + +

    ◆ done() [1/2]

    + +
    +
    +
    +template<class T , Device D>
    + + + + + +
    + + + + + + + + +
    void dlaf::matrix::Matrix< const T, D >::done (const GlobalTileIndex & index)
    +     		+inlinenoexcept
    +
    +

    Mark the tile at index as done

    +

    Marking a tile as done means it can no longer be accessed. Marking a tile as done also disallows creation of sub pipelines from the full matrix.

    + +
    +
    + +

    ◆ done() [2/2]

    + +
    +
    +
    +template<class T , Device D>
    + + + + + +
    + + + + + + + + +
    void dlaf::matrix::Matrix< const T, D >::done (const LocalTileIndex & index)
    +     		+inlinenoexcept
    +
    +

    Mark the tile at index as done

    +

    Marking a tile as done means it can no longer be accessed. Marking a tile as done also disallows creation of sub pipelines from the full matrix.

    + +
    +

    ◆ read() [1/2]

    @@ -300,6 +371,68 @@

    Tile with local index index.

    Precondition
    index.isIn(distribution().localNrTiles()).
    + + + +

    ◆ retiledSubPipelineConst()

    + +
    +
    +
    +template<class T , Device D>
    + + + + + +
    + + + + + + + + +
    Matrix dlaf::matrix::Matrix< const T, D >::retiledSubPipelineConst (const LocalTileSize & tiles_per_block)
    +     		+inline
    +
    +

    Create a sub-pipelined, retiled matrix which can be accessed thread-safely with respect to the original matrix

    +

    All accesses to the sub-pipelined matrix are sequenced after previous accesses and before later accesses to the original matrix, independently of when tiles are accessed in the sub-pipelined matrix.

    +
    Precondition
    blockSize() is divisible by tiles_per_block
    +
    +blockSize() == baseTileSize()
    + +
    +
    + +

    ◆ subPipelineConst()

    + +
    +
    +
    +template<class T , Device D>
    + + + + + +
    + + + + + + + +
    Matrix dlaf::matrix::Matrix< const T, D >::subPipelineConst ()
    +     		+inline
    +
    +

    Create a sub-pipelined matrix which can be accessed thread-safely with respect to the original matrix

    +

    All accesses to the sub-pipelined matrix are sequenced after previous accesses and before later accesses to the original matrix, independently of when tiles are accessed in the sub-pipelined matrix.

    +
    diff --git a/master/classdlaf_1_1matrix_1_1_retiled_matrix-members.html b/master/classdlaf_1_1matrix_1_1_retiled_matrix-members.html deleted file mode 100644 index cde2bec303..0000000000 --- a/master/classdlaf_1_1matrix_1_1_retiled_matrix-members.html +++ /dev/null @@ -1,116 +0,0 @@ - - - - - - - -DLAF: Member List - - - - - - - - - -
    -
    - - - - - - -
    -
    DLAF -
    -
    -
    - - - - - - - - -
    -
    - - -
    - -
    - - -
    -
    -
    -
    dlaf::matrix::RetiledMatrix< T, D > Member List
    -
    -
    - -

    This is the complete list of members for dlaf::matrix::RetiledMatrix< T, D >, including all inherited members.

    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
    baseTileSize() const noexceptdlaf::matrix::internal::MatrixBaseinline
    blockSize() const noexceptdlaf::matrix::internal::MatrixBaseinline
    commGridSize() const noexceptdlaf::matrix::internal::MatrixBaseinline
    ConstTileType typedef (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >
    device (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >static
    distribution() const noexceptdlaf::matrix::internal::MatrixBaseinline
    done(const LocalTileIndex &index) noexcept (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >inline
    done(const GlobalTileIndex &index) noexcept (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >inline
    ElementType typedef (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >
    MatrixBase(Distribution distribution) (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline
    MatrixBase(const Distribution &distribution, const LocalTileSize &tiles_per_block) (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline
    MatrixBase(const MatrixBase &rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBase
    MatrixBase(MatrixBase &&rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseprotected
    nrTiles() const noexceptdlaf::matrix::internal::MatrixBaseinline
    operator=(const RetiledMatrix &rhs)=delete (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >
    operator=(RetiledMatrix &&rhs)=default (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >
    operator=(const MatrixBase &rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBase
    operator=(MatrixBase &&rhs)=default (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseprotected
    rankGlobalTile(const GlobalTileIndex &global_tile) const noexceptdlaf::matrix::internal::MatrixBaseinline
    rankIndex() const noexceptdlaf::matrix::internal::MatrixBaseinline
    read(const LocalTileIndex &index) noexceptdlaf::matrix::RetiledMatrix< T, D >inline
    read(const GlobalTileIndex &index) noexceptdlaf::matrix::RetiledMatrix< T, D >inline
    readwrite(const LocalTileIndex &index) noexceptdlaf::matrix::RetiledMatrix< T, D >inline
    readwrite(const GlobalTileIndex &index) noexceptdlaf::matrix::RetiledMatrix< T, D >inline
    RetiledMatrix(Matrix< T, D > &mat, const LocalTileSize &tiles_per_block)dlaf::matrix::RetiledMatrix< T, D >inline
    RetiledMatrix(const RetiledMatrix &rhs)=delete (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >
    RetiledMatrix(RetiledMatrix &&rhs)=default (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >
    size() const noexceptdlaf::matrix::internal::MatrixBaseinline
    sourceRankIndex() const noexcept (defined in dlaf::matrix::internal::MatrixBase)dlaf::matrix::internal::MatrixBaseinline
    tile_managers_ (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >protected
    TileDataType typedef (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >
    tileLinearIndex(const LocalTileIndex &index) const noexceptdlaf::matrix::internal::MatrixBaseinlineprotected
    tileSize(const GlobalTileIndex &index) const noexceptdlaf::matrix::internal::MatrixBaseinline
    TileType typedef (defined in dlaf::matrix::RetiledMatrix< T, D >)dlaf::matrix::RetiledMatrix< T, D >
    waitLocalTiles() noexceptdlaf::matrix::RetiledMatrix< T, D >inline
    - -
    -Generated by doxygen 1.9.1 -
    - - diff --git a/master/classdlaf_1_1matrix_1_1_retiled_matrix.html b/master/classdlaf_1_1matrix_1_1_retiled_matrix.html deleted file mode 100644 index c4b62fb410..0000000000 --- a/master/classdlaf_1_1matrix_1_1_retiled_matrix.html +++ /dev/null @@ -1,421 +0,0 @@ - - - - - - - -DLAF: dlaf::matrix::RetiledMatrix< T, D > Class Template Reference - - - - - - - - - -
    -
    - - - - - - -
    -
    DLAF -
    -
    -
    - - - - - - - - -
    -
    - - -
    - -
    - - -
    -
    -
    -Public Types | -Public Member Functions | -Static Public Attributes | -Protected Attributes | -List of all members
    -
    -
    dlaf::matrix::RetiledMatrix< T, D > Class Template Reference
    -
    -
    - -

    #include <retiled_matrix.h>

    -
    -Inheritance diagram for dlaf::matrix::RetiledMatrix< T, D >:
    -
    -
    - - -dlaf::matrix::internal::MatrixBase - -
    - - - - - - - - - - -

    -Public Types

    -using ElementType = T
     
    -using TileType = Tile< ElementType, D >
     
    -using ConstTileType = Tile< const ElementType, D >
     
    -using TileDataType = internal::TileData< const ElementType, D >
     
    - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

    -Public Member Functions

     RetiledMatrix (Matrix< T, D > &mat, const LocalTileSize &tiles_per_block)
     
    - RetiledMatrix (const RetiledMatrix &rhs)=delete
     
    - RetiledMatrix (RetiledMatrix &&rhs)=default
     
    -RetiledMatrix & operator= (const RetiledMatrix &rhs)=delete
     
    -RetiledMatrix & operator= (RetiledMatrix &&rhs)=default
     
    ReadOnlyTileSender< T, D > read (const LocalTileIndex &index) noexcept
     
    ReadOnlyTileSender< T, D > read (const GlobalTileIndex &index) noexcept
     
    void waitLocalTiles () noexcept
     
    ReadWriteTileSender< T, D > readwrite (const LocalTileIndex &index) noexcept
     
    ReadWriteTileSender< T, D > readwrite (const GlobalTileIndex &index) noexcept
     
    -void done (const LocalTileIndex &index) noexcept
     
    -void done (const GlobalTileIndex &index) noexcept
     
    - Public Member Functions inherited from dlaf::matrix::internal::MatrixBase
    - MatrixBase (Distribution distribution)
     
    - MatrixBase (const Distribution &distribution, const LocalTileSize &tiles_per_block)
     
    - MatrixBase (const MatrixBase &rhs)=default
     
    -MatrixBase & operator= (const MatrixBase &rhs)=default
     
    -const GlobalElementSize & size () const noexcept
     Returns the global size in elements of the matrix.
     
    -const TileElementSize & blockSize () const noexcept
     Returns the block size of the matrix.
     
    -const TileElementSize & baseTileSize () const noexcept
     Returns the complete tile size of the matrix.
     
    -const GlobalTileSize & nrTiles () const noexcept
     Returns the number of tiles of the global matrix (2D size).
     
    -const comm::Index2D & rankIndex () const noexcept
     Returns the id associated to the matrix of this rank.
     
    -const comm::Index2D & sourceRankIndex () const noexcept
     
    -const comm::Size2D & commGridSize () const noexcept
     Returns the size of the communicator grid associated to the matrix.
     
    comm::Index2D rankGlobalTile (const GlobalTileIndex &global_tile) const noexcept
     
    -const matrix::Distribution & distribution () const noexcept
     Returns the distribution of the matrix.
     
    -TileElementSize tileSize (const GlobalTileIndex &index) const noexcept
     Returns the size of the Tile with global index index.
     
    - - - -

    -Static Public Attributes

    -static constexpr Device device = D
     
    - - - -

    -Protected Attributes

    -std::vector< internal::TilePipeline< T, D > > tile_managers_
     
    - - - - - - - - -

    -Additional Inherited Members

    - Protected Member Functions inherited from dlaf::matrix::internal::MatrixBase
    - MatrixBase (MatrixBase &&rhs)=default
     
    -MatrixBase & operator= (MatrixBase &&rhs)=default
     
    std::size_t tileLinearIndex (const LocalTileIndex &index) const noexcept
     
    -

    Detailed Description

    -

    template<class T, Device D>
    -class dlaf::matrix::RetiledMatrix< T, D >

    - -

    A RetiledMatrix object represents a collection of tiles which contain all the elements of a matrix. It is constructed on top of a Matrix and allows to divide the distribution block in multiple tiles.

    -

    The tiles are distributed according to a distribution (see Matrix::distribution()), therefore some tiles are stored locally on this rank, while the others are available on other ranks.

    -

    Constructor & Destructor Documentation

    - -

    ◆ RetiledMatrix()

    - -
    -
    -
    -template<class T , Device D>
    - - - - - -
    - - - - - - - - - - - - - - - - - - -
    dlaf::matrix::RetiledMatrix< T, D >::RetiledMatrix (Matrix< T, D > & mat,
    const LocalTileSize & tiles_per_block 
    )
    -     		-inline
    -
    -
    Precondition
    mat.blockSize() is divisible by tiles_per_block.
    - -
    -
    -

    Member Function Documentation

    - -

    ◆ read() [1/2]

    - -
    -
    -
    -template<class T , Device D>
    - - - - - -
    - - - - - - - - -
    ReadOnlyTileSender<T, D> dlaf::matrix::RetiledMatrix< T, D >::read (const GlobalTileIndex & index)
    -     		-inlinenoexcept
    -
    -

    Returns a read-only sender of the Tile with global index index.

    -
    Precondition
    the global tile is stored in the current process,
    -
    -index.isIn(globalNrTiles()).
    - -
    -
    - -

    ◆ read() [2/2]

    - -
    -
    -
    -template<class T , Device D>
    - - - - - -
    - - - - - - - - -
    ReadOnlyTileSender<T, D> dlaf::matrix::RetiledMatrix< T, D >::read (const LocalTileIndex & index)
    -     		-inlinenoexcept
    -
    -

    Returns a read-only sender of the Tile with local index index.

    -
    Precondition
    index.isIn(distribution().localNrTiles()).
    - -
    -
    - -

    ◆ readwrite() [1/2]

    - -
    -
    -
    -template<class T , Device D>
    - - - - - -
    - - - - - - - - -
    ReadWriteTileSender<T, D> dlaf::matrix::RetiledMatrix< T, D >::readwrite (const GlobalTileIndex & index)
    -     		-inlinenoexcept
    -
    -

    Returns a sender of the Tile with global index index.

    -
    Precondition
    the global tile is stored in the current process,
    -
    -index.isIn(globalNrTiles()).
    - -
    -
    - -

    ◆ readwrite() [2/2]

    - -
    -
    -
    -template<class T , Device D>
    - - - - - -
    - - - - - - - - -
    ReadWriteTileSender<T, D> dlaf::matrix::RetiledMatrix< T, D >::readwrite (const LocalTileIndex & index)
    -     		-inlinenoexcept
    -
    -

    Returns a sender of the Tile with local index index.

    -
    Precondition
    index.isIn(distribution().localNrTiles()).
    - -
    -
    - -

    ◆ waitLocalTiles()

    - -
    -
    -
    -template<class T , Device D>
    - - - - - -
    - - - - - - - -
    void dlaf::matrix::RetiledMatrix< T, D >::waitLocalTiles ()
    -     		-inlinenoexcept
    -
    -

    Synchronization barrier for all local tiles in the matrix

    -

    This blocking call does not return until all operations, i.e. both RO and RW, involving any of the locally available tiles are completed.

    - -
    -
    -
    The documentation for this class was generated from the following file:
      -
    • /home/runner/work/DLA-Future/DLA-Future/include/dlaf/matrix/retiled_matrix.h
      • -
    -
    - -
    -Generated by doxygen 1.9.1 -
    - - diff --git a/master/classdlaf_1_1matrix_1_1_retiled_matrix.png b/master/classdlaf_1_1matrix_1_1_retiled_matrix.png deleted file mode 100644 index 317ef1cf17..0000000000 Binary files a/master/classdlaf_1_1matrix_1_1_retiled_matrix.png and /dev/null differ diff --git a/master/classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html b/master/classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html index bb8d3273d2..9af618856c 100644 --- a/master/classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html +++ b/master/classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html @@ -80,8 +80,7 @@
    -dlaf::matrix::Matrix< const T, D > -dlaf::matrix::RetiledMatrix< T, D > +dlaf::matrix::Matrix< const T, D >
    diff --git a/master/classdlaf_1_1matrix_1_1internal_1_1_matrix_base.png b/master/classdlaf_1_1matrix_1_1internal_1_1_matrix_base.png index d1641a0d9b..0c7fbbc175 100644 Binary files a/master/classdlaf_1_1matrix_1_1internal_1_1_matrix_base.png and b/master/classdlaf_1_1matrix_1_1internal_1_1_matrix_base.png differ diff --git a/master/classes.html b/master/classes.html index 8262688254..1ccedec598 100644 --- a/master/classes.html +++ b/master/classes.html @@ -111,7 +111,7 @@
    QR (dlaf::factorization::internal)
    QR_Tfactor (dlaf::factorization::internal)
    R
    -
    ReductionToBand (dlaf::eigensolver::internal)
    RetiledMatrix (dlaf::matrix)
    RoundRobin (dlaf::common)
    +
    ReductionToBand (dlaf::eigensolver::internal)
    RoundRobin (dlaf::common)
    S
    ScopedInitializer (dlaf)
    SenderSingleValueTypeImpl (dlaf::internal)
    SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::read > > > > (dlaf::internal)
    SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::readwrite > > > > (dlaf::internal)
    SenderSingleValueTypeImpl< TypeList< TypeList< std::reference_wrapper< T > > > > (dlaf::internal)
    SenderSingleValueTypeImpl< TypeList< TypeList< T > > > (dlaf::internal)
    SingleThreadedBlasScope (dlaf::common::internal)
    Size2D (dlaf::common)
    source_location (dlaf::common::internal)
    SubMatrixView (dlaf::matrix)
    SubPanelView (dlaf::matrix)
    Matrix< const T, D >::SubPipelineTag (dlaf::matrix)
    SubTileSpec (dlaf::matrix)
    SweepWorker (dlaf::eigensolver::internal)
    SweepWorkerDist (dlaf::eigensolver::internal)
    diff --git a/master/copy_8h_source.html b/master/copy_8h_source.html index e97f2399b7..6d6a219a23 100644 --- a/master/copy_8h_source.html +++ b/master/copy_8h_source.html @@ -131,7 +131,7 @@
    dlaf::internal::Policy
    Definition: policy.h:24
    -
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
    Definition: matrix.h:121
    +
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
    Definition: matrix.h:122
    copy_tile.h
    diff --git a/master/dir_21834082df7a318c018c9cc963be86ec.html b/master/dir_21834082df7a318c018c9cc963be86ec.html index bdf58ac8c6..d213edd02b 100644 --- a/master/dir_21834082df7a318c018c9cc963be86ec.html +++ b/master/dir_21834082df7a318c018c9cc963be86ec.html @@ -99,8 +99,6 @@ - - diff --git a/master/eigensolver_2bt__band__to__tridiag_2impl_8h_source.html b/master/eigensolver_2bt__band__to__tridiag_2impl_8h_source.html index 9766a50f56..e8ab1d2087 100644 --- a/master/eigensolver_2bt__band__to__tridiag_2impl_8h_source.html +++ b/master/eigensolver_2bt__band__to__tridiag_2impl_8h_source.html @@ -110,966 +110,965 @@
    39 #include <dlaf/matrix/index.h>
    40 #include <dlaf/matrix/matrix.h>
    41 #include <dlaf/matrix/panel.h>
    -
    42 #include <dlaf/matrix/retiled_matrix.h>
    -
    43 #include <dlaf/matrix/tile.h>
    -
    44 #include <dlaf/sender/policy.h>
    -
    45 #include <dlaf/sender/traits.h>
    -
    46 #include <dlaf/sender/transform.h>
    -
    47 #include <dlaf/sender/when_all_lift.h>
    -
    48 #include <dlaf/tune.h>
    -
    49 #include <dlaf/types.h>
    -
    50 #include <dlaf/util_math.h>
    -
    51 #include <dlaf/util_matrix.h>
    -
    52 
    -
    53 namespace dlaf::eigensolver::internal {
    -
    54 
    -
    55 namespace bt_tridiag {
    -
    56 
    -
    57 template <class T>
    -
    58 matrix::Tile<T, Device::CPU> setupVWellFormed(const SizeType b,
    -
    59  const matrix::Tile<const T, Device::CPU>& tile_hh,
    -
    60  matrix::Tile<T, Device::CPU> tile_v) {
    -
    61  using lapack::lacpy;
    -
    62  using lapack::laset;
    -
    63 
    -
    64  common::internal::SingleThreadedBlasScope single;
    -
    65 
    -
    66  // Note: the size of of tile_hh and tile_v embeds a relevant information about the number of
    -
    67  // reflecotrs and their max size. This will be exploited to correctly setup the well formed
    -
    68  // tile with reflectors in place as they will be applied.
    -
    69  const auto k = tile_v.size().cols();
    -
    70 
    -
    71  // copy from compact representation reflector values (the first component set to 1 is not there)
    -
    72  for (SizeType j = 0; j < k; ++j) {
    -
    73  const auto compact_refl_size =
    -
    74  std::min<SizeType>(tile_v.size().rows() - (1 + j), tile_hh.size().rows() - 1);
    -
    75 
    -
    76  // this is needed because of complex last reflector (i.e. just 1 element long)
    -
    77  if (compact_refl_size == 0)
    -
    78  continue;
    -
    79 
    -
    80  lacpy(blas::Uplo::General, compact_refl_size, 1, tile_hh.ptr({1, j}), tile_hh.ld(),
    -
    81  tile_v.ptr({1 + j, j}), tile_v.ld());
    -
    82  }
    -
    83 
    -
    84  // Note:
    -
    85  // In addition to setting the diagonal to 1 for missing first components, here it zeros out
    -
    86  // both the upper and the lower part. Indeed due to the skewed shape, reflectors do not occupy
    -
    87  // the full tile height, and V should be fully well-formed because the next triangular
    -
    88  // multiplication, i.e. `V . T`, and the gemm `V* . E`, will use V as a general matrix.
    -
    89  laset(blas::Uplo::Upper, tile_v.size().rows(), k, T(0), T(1), tile_v.ptr({0, 0}), tile_v.ld());
    -
    90 
    -
    91  if (tile_v.size().rows() > b)
    -
    92  laset(blas::Uplo::Lower, tile_v.size().rows() - b, k - 1, T(0), T(0), tile_v.ptr({b, 0}),
    -
    93  tile_v.ld());
    -
    94 
    -
    95  return tile_v;
    -
    96 }
    -
    97 
    -
    98 template <class T>
    -
    99 void computeTFactor(const matrix::Tile<const T, Device::CPU>& tile_taus,
    -
    100  const matrix::Tile<const T, Device::CPU>& tile_v,
    -
    101  const matrix::Tile<T, Device::CPU>& tile_t) {
    -
    102  using namespace lapack;
    -
    103 
    -
    104  common::internal::SingleThreadedBlasScope single;
    -
    105 
    -
    106  // taus have to be extracted from the compact form (i.e. first row of the input tile)
    -
    107  std::vector<T> taus;
    -
    108  taus.resize(to_sizet(tile_v.size().cols()));
    -
    109  for (SizeType i = 0; i < to_SizeType(taus.size()); ++i)
    -
    110  taus[to_sizet(i)] = tile_taus({0, i});
    -
    111 
    -
    112  const auto n = tile_v.size().rows();
    -
    113  const auto k = tile_v.size().cols();
    -
    114  larft(Direction::Forward, StoreV::Columnwise, n, k, tile_v.ptr(), tile_v.ld(), taus.data(),
    -
    115  tile_t.ptr(), tile_t.ld());
    -
    116 }
    -
    117 
    -
    118 template <class T>
    -
    119 std::tuple<matrix::Tile<T, Device::CPU>, matrix::Tile<T, Device::CPU>> computeVT(
    -
    120  const SizeType b, const matrix::Tile<const T, Device::CPU>& tile_hh, const SizeType hhr_nb,
    -
    121  matrix::Tile<T, Device::CPU> tile_v, matrix::Tile<T, Device::CPU> tile_t) {
    -
    122  auto tile_v2 = setupVWellFormed(b, tile_hh, std::move(tile_v));
    -
    123  for (SizeType j = 0; j < tile_v2.size().cols(); j += hhr_nb) {
    -
    124  const SizeType jb = std::min(hhr_nb, tile_v2.size().cols() - j);
    -
    125  const SizeType ib = std::min(jb + b - 1, tile_v2.size().rows() - j);
    -
    126  auto subtile_t = tile_t.subTileReference({{j, j}, {jb, jb}});
    -
    127  auto subtile_hh = tile_hh.subTileReference({{0, j}, {1, jb}});
    -
    128  auto subtile_v = tile_v2.subTileReference({{j, j}, {ib, jb}});
    -
    129  computeTFactor(subtile_hh, subtile_v, subtile_t);
    -
    130  }
    -
    131  return std::make_tuple(std::move(tile_v2), std::move(tile_t));
    -
    132 }
    -
    133 
    -
    134 template <class T>
    -
    135 std::tuple<matrix::Tile<T, Device::CPU>, matrix::Tile<T, Device::CPU>> computeVW(
    -
    136  const SizeType b, const matrix::Tile<const T, Device::CPU>& tile_hh, const SizeType hhr_nb,
    -
    137  matrix::Tile<T, Device::CPU> tile_v, matrix::Tile<T, Device::CPU> tile_t,
    -
    138  matrix::Tile<T, Device::CPU> tile_w) {
    -
    139  using namespace blas;
    -
    140 
    -
    141  auto [tile_v2, tile_t2] = computeVT(b, tile_hh, hhr_nb, std::move(tile_v), std::move(tile_t));
    -
    142 
    -
    143  for (SizeType j = 0; j < tile_v2.size().cols(); j += hhr_nb) {
    -
    144  const SizeType jb = std::min(hhr_nb, tile_v2.size().cols() - j);
    -
    145  const SizeType ib = std::min(jb + b - 1, tile_v2.size().rows() - j);
    -
    146  auto subtile_t = tile_t2.subTileReference({{j, j}, {jb, jb}});
    -
    147  auto subtile_v = tile_v2.subTileReference({{j, j}, {ib, jb}});
    -
    148  auto subtile_w = tile_w.subTileReference({{j, j}, {ib, jb}});
    -
    149 
    -
    150  dlaf::tile::internal::trmm3(Side::Right, Uplo::Upper, Op::NoTrans, Diag::NonUnit, T(1), subtile_t,
    -
    151  subtile_v, subtile_w);
    -
    152  }
    -
    153  return std::make_tuple(std::move(tile_v2), std::move(tile_w));
    -
    154 }
    -
    155 
    -
    156 template <class Tile, class CTile>
    -
    157 std::tuple<CTile, CTile, Tile, Tile> applyHHToSingleTileRowSubtileHelper( //
    -
    158  const SizeType j, const SizeType jb, const CTile& tile_v, const CTile& tile_w, const Tile& tile_w2,
    -
    159  const Tile& tile_e) {
    -
    160  DLAF_ASSERT_HEAVY(tile_v.size() == tile_w.size(), tile_v, tile_w);
    -
    161  DLAF_ASSERT_HEAVY(tile_e.size().rows() - 1 == tile_v.size().rows(), tile_e, tile_v);
    -
    162  DLAF_ASSERT_HEAVY(tile_e.size().cols() == tile_w2.size().cols(), tile_e, tile_w2);
    -
    163 
    -
    164  const SizeType ib = tile_v.size().rows() - j;
    -
    165  auto subtile_v = tile_v.subTileReference({{j, j}, {ib, jb}});
    -
    166  auto subtile_w = tile_w.subTileReference({{j, j}, {ib, jb}});
    -
    167  auto subtile_w2 = tile_w2.subTileReference({{0, 0}, {jb, tile_w2.size().cols()}});
    -
    168  auto subtile_e = tile_e.subTileReference({{j + 1, 0}, tile_e.size() - TileElementSize{j + 1, 0}});
    -
    169 
    -
    170  return {std::move(subtile_v), std::move(subtile_w), std::move(subtile_w2), std::move(subtile_e)};
    -
    171 }
    -
    172 
    -
    173 template <Backend B, class T>
    -
    174 struct ApplyHHToSingleTileRow;
    -
    175 
    -
    176 template <class T>
    -
    177 struct ApplyHHToSingleTileRow<Backend::MC, T> {
    -
    178  void operator()(const SizeType hhr_nb, const matrix::Tile<const T, Device::CPU>& tile_v,
    -
    179  const matrix::Tile<const T, Device::CPU>& tile_w,
    -
    180  const matrix::Tile<T, Device::CPU>& tile_w2,
    -
    181  const matrix::Tile<T, Device::CPU>& tile_e) {
    -
    182  using namespace blas;
    -
    183  using tile::internal::gemm;
    -
    184 
    -
    185  common::internal::SingleThreadedBlasScope single;
    -
    186 
    -
    187  for (SizeType j = (util::ceilDiv(tile_v.size().cols(), hhr_nb) - 1) * hhr_nb; j >= 0; j -= hhr_nb) {
    -
    188  const SizeType jb = std::min(hhr_nb, tile_v.size().cols() - j);
    -
    189  auto [subtile_v, subtile_w, subtile_w2, subtile_e] =
    -
    190  applyHHToSingleTileRowSubtileHelper(j, jb, tile_v, tile_w, tile_w2, tile_e);
    -
    191 
    -
    192  // W2 = V* . E
    -
    193  gemm(Op::ConjTrans, Op::NoTrans, T(1), subtile_v, subtile_e, T(0), subtile_w2);
    -
    194  // E -= W . W2
    -
    195  gemm(Op::NoTrans, Op::NoTrans, T(-1), subtile_w, subtile_w2, T(1), subtile_e);
    -
    196  }
    -
    197  }
    -
    198 };
    -
    199 
    -
    200 #ifdef DLAF_WITH_GPU
    -
    201 template <class T>
    -
    202 struct ApplyHHToSingleTileRow<Backend::GPU, T> {
    -
    203  void operator()(cublasHandle_t handle, const SizeType hhr_nb,
    -
    204  const matrix::Tile<const T, Device::GPU>& tile_v,
    -
    205  const matrix::Tile<const T, Device::GPU>& tile_w,
    -
    206  const matrix::Tile<T, Device::GPU>& tile_w2,
    -
    207  const matrix::Tile<T, Device::GPU>& tile_e) {
    -
    208  using namespace blas;
    -
    209  using tile::internal::gemm;
    -
    210 
    -
    211  common::internal::SingleThreadedBlasScope single;
    -
    212 
    -
    213  for (SizeType j = (util::ceilDiv(tile_v.size().cols(), hhr_nb) - 1) * hhr_nb; j >= 0; j -= hhr_nb) {
    -
    214  const SizeType jb = std::min(hhr_nb, tile_v.size().cols() - j);
    -
    215  auto [subtile_v, subtile_w, subtile_w2, subtile_e] =
    -
    216  applyHHToSingleTileRowSubtileHelper(j, jb, tile_v, tile_w, tile_w2, tile_e);
    -
    217 
    -
    218  // W2 = V* . E
    -
    219  gemm(handle, Op::ConjTrans, Op::NoTrans, T(1), subtile_v, subtile_e, T(0), subtile_w2);
    -
    220  // E -= W . W2
    -
    221  gemm(handle, Op::NoTrans, Op::NoTrans, T(-1), subtile_w, subtile_w2, T(1), subtile_e);
    -
    222  }
    -
    223  }
    -
    224 };
    -
    225 #endif
    -
    226 
    -
    227 template <class Tile, class CTile>
    -
    228 std::tuple<CTile, CTile, CTile, CTile, Tile, Tile, Tile> applyHHToDoubleTileRowSubtileHelper( //
    -
    229  const SizeType j, const SizeType jb, const CTile& tile_v, const CTile& tile_w, const Tile& tile_w2,
    -
    230  const Tile& tile_e_top, const Tile& tile_e_bottom) {
    -
    231  DLAF_ASSERT_HEAVY(tile_v.size() == tile_w.size(), tile_v, tile_w);
    -
    232  DLAF_ASSERT_HEAVY(tile_e_top.size().rows() + tile_e_bottom.size().rows() - 1 == tile_v.size().rows(),
    -
    233  tile_e_top, tile_e_bottom, tile_v);
    -
    234  DLAF_ASSERT_HEAVY(tile_e_top.size().cols() == tile_w2.size().cols(), tile_e_top, tile_w2);
    -
    235  DLAF_ASSERT_HEAVY(tile_e_bottom.size().cols() == tile_w2.size().cols(), tile_e_bottom, tile_w2);
    -
    236 
    -
    237  auto subtile_e_top =
    -
    238  tile_e_top.subTileReference({{j + 1, 0}, tile_e_top.size() - TileElementSize{j + 1, 0}});
    -
    239  auto subtile_e_bottom = tile_e_bottom.subTileReference(
    -
    240  {{0, 0},
    -
    241  TileElementSize{std::min(tile_e_bottom.size().rows(), j + jb), tile_e_bottom.size().cols()}});
    -
    242 
    -
    243  matrix::SubTileSpec spec_top{{j, j}, {subtile_e_top.size().rows(), jb}};
    -
    244  matrix::SubTileSpec spec_bottom{{tile_e_top.size().rows() - 1, j},
    -
    245  {subtile_e_bottom.size().rows(), jb}};
    -
    246 
    -
    247  auto subtile_v_top = tile_v.subTileReference(spec_top);
    -
    248  auto subtile_v_bottom = tile_v.subTileReference(spec_bottom);
    -
    249  auto subtile_w_top = tile_w.subTileReference(spec_top);
    -
    250  auto subtile_w_bottom = tile_w.subTileReference(spec_bottom);
    -
    251  auto subtile_w2 = tile_w2.subTileReference({{0, 0}, {jb, tile_w2.size().cols()}});
    -
    252 
    -
    253  return {std::move(subtile_v_top), std::move(subtile_v_bottom), std::move(subtile_w_top),
    -
    254  std::move(subtile_w_bottom), std::move(subtile_w2), std::move(subtile_e_top),
    -
    255  std::move(subtile_e_bottom)};
    -
    256 }
    -
    257 
    -
    258 template <Backend B, class T>
    -
    259 struct ApplyHHToDoubleTileRow;
    -
    260 
    -
    261 template <class T>
    -
    262 struct ApplyHHToDoubleTileRow<Backend::MC, T> {
    -
    263  void operator()(const SizeType hhr_nb, const matrix::Tile<const T, Device::CPU>& tile_v,
    -
    264  const matrix::Tile<const T, Device::CPU>& tile_w,
    -
    265  const matrix::Tile<T, Device::CPU>& tile_w2,
    -
    266  const matrix::Tile<T, Device::CPU>& tile_e_top,
    -
    267  const matrix::Tile<T, Device::CPU>& tile_e_bottom) {
    -
    268  using namespace blas;
    -
    269  using tile::internal::gemm;
    -
    270 
    -
    271  common::internal::SingleThreadedBlasScope single;
    -
    272 
    -
    273  for (SizeType j = (util::ceilDiv(tile_v.size().cols(), hhr_nb) - 1) * hhr_nb; j >= 0; j -= hhr_nb) {
    -
    274  const SizeType jb = std::min(hhr_nb, tile_v.size().cols() - j);
    -
    275  auto [subtile_v_top, subtile_v_bottom, subtile_w_top, subtile_w_bottom, subtile_w2, subtile_e_top,
    -
    276  subtile_e_bottom] =
    -
    277  applyHHToDoubleTileRowSubtileHelper(j, jb, tile_v, tile_w, tile_w2, tile_e_top, tile_e_bottom);
    -
    278 
    -
    279  // W2 = V* . E
    -
    280  gemm(Op::ConjTrans, Op::NoTrans, T(1), subtile_v_top, subtile_e_top, T(0), subtile_w2);
    -
    281  gemm(Op::ConjTrans, Op::NoTrans, T(1), subtile_v_bottom, subtile_e_bottom, T(1), subtile_w2);
    -
    282  // E -= W . W2
    -
    283  gemm(Op::NoTrans, Op::NoTrans, T(-1), subtile_w_top, subtile_w2, T(1), subtile_e_top);
    -
    284  gemm(Op::NoTrans, Op::NoTrans, T(-1), subtile_w_bottom, subtile_w2, T(1), subtile_e_bottom);
    -
    285  }
    -
    286  }
    -
    287 };
    -
    288 
    -
    289 #ifdef DLAF_WITH_GPU
    -
    290 template <class T>
    -
    291 struct ApplyHHToDoubleTileRow<Backend::GPU, T> {
    -
    292  void operator()(cublasHandle_t handle, const SizeType hhr_nb,
    -
    293  const matrix::Tile<const T, Device::GPU>& tile_v,
    -
    294  const matrix::Tile<const T, Device::GPU>& tile_w,
    -
    295  const matrix::Tile<T, Device::GPU>& tile_w2,
    -
    296  const matrix::Tile<T, Device::GPU>& tile_e_top,
    -
    297  const matrix::Tile<T, Device::GPU>& tile_e_bottom) {
    -
    298  using namespace blas;
    -
    299  using tile::internal::gemm;
    -
    300 
    -
    301  common::internal::SingleThreadedBlasScope single;
    -
    302 
    -
    303  for (SizeType j = (util::ceilDiv(tile_v.size().cols(), hhr_nb) - 1) * hhr_nb; j >= 0; j -= hhr_nb) {
    -
    304  const SizeType jb = std::min(hhr_nb, tile_v.size().cols() - j);
    -
    305  auto [subtile_v_top, subtile_v_bottom, subtile_w_top, subtile_w_bottom, subtile_w2, subtile_e_top,
    -
    306  subtile_e_bottom] =
    -
    307  applyHHToDoubleTileRowSubtileHelper(j, jb, tile_v, tile_w, tile_w2, tile_e_top, tile_e_bottom);
    -
    308 
    -
    309  // W2 = V* . E
    -
    310  gemm(handle, Op::ConjTrans, Op::NoTrans, T(1), subtile_v_top, subtile_e_top, T(0), subtile_w2);
    -
    311  gemm(handle, Op::ConjTrans, Op::NoTrans, T(1), subtile_v_bottom, subtile_e_bottom, T(1),
    -
    312  subtile_w2);
    -
    313  // E -= W . W2
    -
    314  gemm(handle, Op::NoTrans, Op::NoTrans, T(-1), subtile_w_top, subtile_w2, T(1), subtile_e_top);
    -
    315  gemm(handle, Op::NoTrans, Op::NoTrans, T(-1), subtile_w_bottom, subtile_w2, T(1),
    -
    316  subtile_e_bottom);
    -
    317  }
    -
    318  }
    -
    319 };
    -
    320 #endif
    -
    321 
    -
    322 struct TileAccessHelper {
    -
    323  TileAccessHelper(const SizeType b, const SizeType nrefls, const matrix::Distribution& dist_hh,
    -
    324  const matrix::Distribution& dist_e, const GlobalElementIndex offset)
    -
    325  : nrefls_(nrefls), input_spec_{dist_hh.tileElementIndex(offset),
    -
    326  {std::min(b, dist_hh.size().rows() - offset.row()),
    -
    327  std::min(b, dist_hh.size().cols() - offset.col())}},
    -
    328  index_e_(dist_e.globalTileFromGlobalElement<Coord::Row>(offset.row())) {
    -
    329  // Note:
    -
    330  // Next logic is about detecting the available application space for reflectors, which once
    -
    331  // extracted they expands to a matrix with 2 * b - 1 height.
    -
    332  //
    -
    333  // There are two main scenarios:
    -
    334  // - reflectors involves rows of a single tile;
    -
    335  // - reflectors involves rows across two different tiles.
    -
    336 
    -
    337  rows_v_ = std::min(2 * b, dist_e.size().rows() - offset.row()) - 1;
    -
    338  rows_v_top_ = std::min(rows_v_, b - 1);
    -
    339 
    -
    340  // Note:
    -
    341  // In general, keep in mind that the first row of the top tile_e should not be touched.
    -
    342  // The structs ApplyHHTo... take care of it.
    -
    343  //
    -
    344  // e.g. b = 4
    -
    345  // reflectors matrix
    -
    346  // X X X X
    -
    347  // 1 0 0 0 X X X X
    -
    348  // a 1 0 0 X X X X
    -
    349  // a b 1 0 X X X X
    -
    350  // -------
    -
    351  // a b c 1 Y Y Y Y
    -
    352  // 0 b c d Y Y Y Y
    -
    353  // 0 0 c d Y Y Y Y
    -
    354  // 0 0 0 d Y Y Y Y
    -
    355  //
    -
    356  // From the drawing above, it is possible to see the dashed tile separation between X and Y,
    -
    357  // and how the reflectors on the left are going to be applied. In particular, the first row of
    -
    358  // the upper tile is not affected.
    -
    359 
    -
    360  // Note:
    -
    361  // A single tile is involved only if it is the last row tile, so by construction reflectors will be
    -
    362  // applied to a single tile;
    -
    363  across_tiles_ = !(index_e_ == dist_e.nrTiles().rows() - 1);
    -
    364 
    -
    365  across_ranks_ = across_tiles_ && (dist_e.rankGlobalTile<Coord::Row>(index_e_) !=
    -
    366  dist_e.rankGlobalTile<Coord::Row>(index_e_ + 1));
    -
    367  }
    -
    368 
    -
    369  // Return true if the application of Householder reflectors involves multiple tiles
    -
    370  bool affectsMultipleTiles() const noexcept {
    -
    371  return across_tiles_;
    -
    372  }
    -
    373 
    -
    374  // Return true if the application of Householder reflectors involves multiple ranks
    -
    375  bool affectsMultipleRanks() const noexcept {
    -
    376  return across_ranks_;
    -
    377  }
    -
    378 
    -
    379  // Return SubTileSpec to use for accessing Householder reflectors in compact form
    -
    380  //
    -
    381  // SubTileSpec returned points to the sub-block in the full tile containing the HH data in compact
    -
    382  // form. If @p reset_origin is true, then the origin component of the SubTileSpec is resetted and
    -
    383  // it will just describe the size of the sub-block containing the Householder reflectors (useful
    -
    384  // for panel access which might not have full-tiles).
    -
    385  matrix::SubTileSpec specHHCompact(const bool reset_origin = false) const noexcept {
    -
    386  if (reset_origin)
    -
    387  return {{0, 0}, input_spec_.size};
    -
    388  return input_spec_;
    -
    389  }
    -
    390 
    -
    391  // Return SubTileSpec to use for accessing Householder reflectors in well formed form
    -
    392  matrix::SubTileSpec specHH() const noexcept {
    -
    393  return {{0, 0}, {rows_v_, nrefls_}};
    -
    394  }
    -
    395 
    -
    396  // Return SubTileSpec to use for accessing Householder reflectors in well formed form
    -
    397  // for the distributed case.
    -
    398  matrix::SubTileSpec specHH(bool is_top_rank) const noexcept {
    -
    399  if (is_top_rank)
    -
    400  return {{0, 0}, {rows_v_top_, nrefls_}};
    -
    401  DLAF_ASSERT_MODERATE(affectsMultipleTiles(), affectsMultipleTiles());
    -
    402  return {{rows_v_top_, 0}, {rows_v_ - rows_v_top_, nrefls_}};
    -
    403  }
    -
    404 
    -
    405  // Return SubTileSpec to use for accessing T factor
    -
    406  matrix::SubTileSpec specT() const noexcept {
    -
    407  return {{0, 0}, {nrefls_, nrefls_}};
    -
    408  }
    -
    409 
    -
    410  matrix::SubTileSpec specW2(const SizeType cols) const noexcept {
    -
    411  return {{0, 0}, {nrefls_, cols}};
    -
    412  }
    -
    413 
    -
    414  GlobalTileIndex topIndexE(const SizeType j) const noexcept {
    -
    415  return {index_e_, j};
    -
    416  }
    -
    417 
    -
    418  GlobalTileIndex bottomIndexE(const SizeType j) const noexcept {
    -
    419  DLAF_ASSERT_MODERATE(affectsMultipleTiles(), affectsMultipleTiles());
    -
    420  return {index_e_ + 1, j};
    -
    421  }
    -
    422 
    -
    423 private:
    -
    424  SizeType nrefls_;
    -
    425  matrix::SubTileSpec input_spec_;
    -
    426  SizeType index_e_;
    -
    427  SizeType rows_v_;
    -
    428  SizeType rows_v_top_;
    -
    429 
    -
    430  bool across_tiles_;
    -
    431  bool across_ranks_;
    -
    432 };
    -
    433 
    -
    446 struct DistIndexing {
    -
    447  DistIndexing(const TileAccessHelper& helper, const matrix::Distribution& dist_hh, const SizeType b,
    -
    448  const GlobalTileIndex& ij, const SizeType& ij_b_row)
    -
    449  : dist_hh(dist_hh), b(b), mb(dist_hh.blockSize().rows()), helper(helper), ij(ij),
    -
    450  ij_b_row(ij_b_row) {
    -
    451  rank = dist_hh.rankIndex();
    -
    452  rankHH = dist_hh.rankGlobalTile(ij);
    -
    453  n_ws_per_block = to_SizeType(static_cast<size_t>(std::ceil(mb / b / 2.0f)) + 1);
    -
    454  }
    -
    455 
    -
    456  comm::IndexT_MPI rankRowPartner() const {
    -
    457  return (rankHH.row() + 1) % dist_hh.commGridSize().rows();
    -
    458  }
    -
    459 
    -
    460  bool isInvolved() const {
    -
    461  const bool isSameRow = rank.row() == rankHH.row();
    -
    462  const bool isPartnerRow = rank.row() == rankRowPartner();
    -
    463  return isSameRow || (isPartnerRow && helper.affectsMultipleRanks());
    -
    464  }
    -
    465 
    -
    466  LocalTileIndex wsIndexHH() const {
    -
    467  const SizeType row = [&]() -> SizeType {
    -
    468  if (rank.row() == rankHH.row()) {
    -
    469  // Note: index starts at 1 (0 is the extra workspace), moreover max half blocks will run in parallel
    -
    470  const SizeType intra_idx = 1 + (ij_b_row % (mb / b)) / 2;
    -
    471  DLAF_ASSERT_HEAVY(intra_idx < n_ws_per_block, intra_idx, n_ws_per_block);
    -
    472  return dist_hh.localTileFromGlobalTile<Coord::Row>(ij.row()) * n_ws_per_block + intra_idx;
    -
    473  }
    -
    474  else {
    -
    475  DLAF_ASSERT_HEAVY(helper.affectsMultipleRanks() && (rank.row() == rankRowPartner()),
    -
    476  helper.affectsMultipleRanks(), rank.row(), rankRowPartner());
    -
    477  return dist_hh.localNrTiles().isEmpty()
    -
    478  ? 0
    -
    479  : dist_hh.localTileFromGlobalTile<Coord::Row>(ij.row() + 1) * n_ws_per_block;
    -
    480  }
    -
    481  }();
    -
    482  return {row, 0};
    -
    483  }
    -
    484 
    -
    485 protected:
    -
    486  matrix::Distribution dist_hh;
    -
    487  SizeType b;
    -
    488  SizeType mb;
    -
    489  SizeType n_ws_per_block;
    -
    490 
    -
    491  TileAccessHelper helper;
    -
    492 
    -
    493  comm::Index2D rank;
    -
    494  comm::Index2D rankHH;
    -
    495 
    -
    496  GlobalTileIndex ij;
    -
    497  SizeType ij_b_row;
    -
    498 };
    -
    499 
    -
    500 template <Backend B, Device D, class T>
    -
    501 struct HHManager;
    -
    502 
    -
    503 template <class T>
    -
    504 struct HHManager<Backend::MC, Device::CPU, T> {
    -
    505  static constexpr Backend B = Backend::MC;
    -
    506  static constexpr Device D = Device::CPU;
    -
    507 
    -
    508  HHManager(const SizeType b, const std::size_t, matrix::Distribution, matrix::Distribution) : b(b) {}
    -
    509 
    -
    510  template <class SenderHH>
    -
    511  auto computeVW(const SizeType nb_apply, const LocalTileIndex ij, const TileAccessHelper& helper,
    -
    512  SenderHH&& tile_hh, matrix::Panel<Coord::Col, T, D>& mat_v,
    -
    513  matrix::Panel<Coord::Col, T, D>& mat_t, matrix::Panel<Coord::Col, T, D>& mat_w) {
    -
    514  namespace ex = pika::execution::experimental;
    -
    515 
    -
    516  return dlaf::internal::whenAllLift(b, std::forward<SenderHH>(tile_hh), nb_apply,
    -
    517  splitTile(mat_v.readwrite(ij), helper.specHH()),
    -
    518  splitTile(mat_t.readwrite(ij), helper.specT()),
    -
    519  splitTile(mat_w.readwrite(ij), helper.specHH())) |
    -
    520  dlaf::internal::transform(dlaf::internal::Policy<Backend::MC>(), bt_tridiag::computeVW<T>) |
    -
    521  ex::split_tuple();
    -
    522  }
    -
    523 
    -
    524 protected:
    -
    525  const SizeType b;
    -
    526 };
    -
    527 
    -
    528 #ifdef DLAF_WITH_GPU
    -
    529 template <class T>
    -
    530 struct HHManager<Backend::GPU, Device::GPU, T> {
    -
    531  static constexpr Backend B = Backend::GPU;
    -
    532  static constexpr Device D = Device::GPU;
    -
    533 
    -
    534  HHManager(const SizeType b, const std::size_t n_workspaces, matrix::Distribution dist_t,
    -
    535  matrix::Distribution dist_w)
    -
    536  : b(b), t_panels_h(n_workspaces, dist_t), w_panels_h(n_workspaces, dist_w) {}
    -
    537 
    -
    538  template <class SenderHH>
    -
    539  auto computeVW(const SizeType hhr_nb, const LocalTileIndex ij, const TileAccessHelper& helper,
    -
    540  SenderHH&& tile_hh, matrix::Panel<Coord::Col, T, D>& mat_v,
    -
    541  matrix::Panel<Coord::Col, T, D>& mat_t, matrix::Panel<Coord::Col, T, D>& mat_w) {
    -
    542  namespace ex = pika::execution::experimental;
    -
    543 
    -
    544  auto& mat_v_h = w_panels_h.nextResource();
    -
    545  auto& mat_t_h = t_panels_h.nextResource();
    -
    546 
    -
    547  const LocalTileIndex ij_t = ij;
    -
    548  const matrix::SubTileSpec t_spec = helper.specT();
    -
    549 
    -
    550  auto [tile_v_h, tile_t_h] =
    -
    551  dlaf::internal::whenAllLift(b, std::forward<SenderHH>(tile_hh), hhr_nb,
    -
    552  splitTile(mat_v_h.readwrite(ij), helper.specHH()),
    -
    553  splitTile(mat_t_h.readwrite(ij_t), t_spec)) |
    -
    554  dlaf::internal::transform(dlaf::internal::Policy<Backend::MC>(), computeVT<T>) |
    -
    555  ex::split_tuple();
    -
    556 
    -
    557  auto copyVTandComputeW =
    -
    558  [b = this->b, hhr_nb](cublasHandle_t handle, const matrix::Tile<const T, Device::CPU>& tile_v_h,
    -
    559  const matrix::Tile<const T, Device::CPU>& tile_t_h,
    -
    560  matrix::Tile<T, Device::GPU>& tile_v, matrix::Tile<T, Device::GPU>& tile_t,
    -
    561  matrix::Tile<T, Device::GPU>& tile_w) {
    -
    562  whip::stream_t stream;
    -
    563  DLAF_GPUBLAS_CHECK_ERROR(cublasGetStream(handle, &stream));
    -
    564 
    -
    565  matrix::internal::copy(tile_v_h, tile_v, stream);
    -
    566  matrix::internal::copy(tile_t_h, tile_t, stream);
    -
    567 
    -
    568  // W = V . T
    -
    569  using namespace blas;
    -
    570  for (SizeType j = 0; j < tile_v_h.size().cols(); j += hhr_nb) {
    -
    571  const SizeType jb = std::min(hhr_nb, tile_v_h.size().cols() - j);
    -
    572  const SizeType ib = std::min(jb + b - 1, tile_v_h.size().rows() - j);
    -
    573  auto subtile_t = tile_t.subTileReference({{j, j}, {jb, jb}});
    -
    574  auto subtile_v = tile_v.subTileReference({{j, j}, {ib, jb}});
    -
    575  auto subtile_w = tile_w.subTileReference({{j, j}, {ib, jb}});
    -
    576 
    -
    577  dlaf::tile::internal::trmm3(handle, Side::Right, Uplo::Upper, Op::NoTrans, Diag::NonUnit,
    -
    578  T(1), subtile_t, subtile_v, subtile_w);
    -
    579  }
    -
    580 
    -
    581  return std::make_tuple(std::move(tile_v), std::move(tile_w));
    -
    582  };
    -
    583 
    -
    584  return ex::when_all(std::move(tile_v_h), std::move(tile_t_h),
    -
    585  splitTile(mat_v.readwrite(ij), helper.specHH()),
    -
    586  splitTile(mat_t.readwrite(ij_t), t_spec),
    -
    587  splitTile(mat_w.readwrite(ij), helper.specHH())) |
    -
    588  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    -
    589  dlaf::internal::Policy<Backend::GPU>(), copyVTandComputeW) |
    -
    590  ex::split_tuple();
    -
    591  }
    -
    592 
    -
    593 protected:
    -
    594  const SizeType b;
    -
    595  common::RoundRobin<matrix::Panel<Coord::Col, T, Device::CPU>> t_panels_h;
    -
    596  common::RoundRobin<matrix::Panel<Coord::Col, T, Device::CPU>> w_panels_h;
    -
    597 };
    -
    598 #endif
    -
    599 }
    -
    600 
    -
    601 template <Backend B, Device D, class T>
    -
    602 void BackTransformationT2B<B, D, T>::call(const SizeType band_size, Matrix<T, D>& mat_e,
    -
    603  Matrix<const T, Device::CPU>& mat_hh) {
    -
    604  using pika::execution::thread_priority;
    -
    605  namespace ex = pika::execution::experimental;
    -
    606 
    -
    607  using common::iterate_range2d;
    -
    608  using common::RoundRobin;
    -
    609  using matrix::Panel;
    -
    610  using namespace bt_tridiag;
    -
    611 
    -
    612  if (mat_hh.size().isEmpty() || mat_e.size().isEmpty())
    -
    613  return;
    -
    614 
    -
    615  // Note: if no householder reflectors are going to be applied (in case of trivial matrix)
    -
    616  if (mat_hh.size().rows() <= (dlaf::isComplex_v<T> ? 1 : 2))
    -
    617  return;
    -
    618 
    -
    619  const SizeType b = band_size;
    -
    620  const SizeType group_size = getTuneParameters().bt_band_to_tridiag_hh_apply_group_size;
    -
    621  const SizeType nsweeps = nrSweeps<T>(mat_hh.size().cols());
    -
    622 
    -
    623  const LocalTileSize tiles_per_block(mat_e.blockSize().rows() / b, 1);
    -
    624  matrix::RetiledMatrix<T, D> mat_e_rt(mat_e, tiles_per_block);
    -
    625 
    -
    626  const auto& dist_hh = mat_hh.distribution();
    -
    627  const auto& dist_e_rt = mat_e_rt.distribution();
    -
    628 
    -
    629  // Note: w_tile_sz can store reflectors as they are actually applied, opposed to how they are
    -
    630  // stored in compact form.
    -
    631  //
    -
    632  // e.g. Given b = 4
    -
    633  //
    -
    634  // compact w_tile_sz
    -
    635  // 1 1 1 1 1 0 0 0
    -
    636  // a b c d a 1 0 0
    -
    637  // a b c d a b 1 0
    -
    638  // a b c d a b c 1
    -
    639  // 0 b c d
    -
    640  // 0 0 c d
    -
    641  // 0 0 0 d
    -
    642  const TileElementSize w_tile_sz(2 * b - 1, b);
    -
    643 
    -
    644  const SizeType dist_w_rows = mat_e_rt.nrTiles().rows() * w_tile_sz.rows();
    -
    645  const matrix::Distribution dist_w({dist_w_rows, b}, w_tile_sz);
    -
    646  const matrix::Distribution dist_t({mat_hh.size().rows(), b}, {b, b});
    -
    647  const matrix::Distribution dist_w2({b, mat_e_rt.size().cols()}, {b, mat_e_rt.blockSize().cols()});
    -
    648 
    -
    649  constexpr std::size_t n_workspaces = 2;
    -
    650  RoundRobin<Panel<Coord::Col, T, D>> t_panels(n_workspaces, dist_t);
    -
    651  RoundRobin<Panel<Coord::Col, T, D>> v_panels(n_workspaces, dist_w);
    -
    652  RoundRobin<Panel<Coord::Col, T, D>> w_panels(n_workspaces, dist_w);
    -
    653  RoundRobin<Panel<Coord::Row, T, D>> w2_panels(n_workspaces, dist_w2);
    -
    654 
    -
    655  HHManager<B, D, T> helperBackend(b, n_workspaces, dist_t, dist_w);
    -
    656 
    -
    657  // Note: sweep are on diagonals, steps are on verticals
    -
    658  const SizeType j_last_sweep = (nsweeps - 1) / b;
    -
    659  for (SizeType j = j_last_sweep; j >= 0; --j) {
    -
    660  auto& mat_t = t_panels.nextResource();
    -
    661  auto& mat_v = v_panels.nextResource();
    -
    662  auto& mat_w = w_panels.nextResource();
    -
    663  auto& mat_w2 = w2_panels.nextResource();
    -
    664 
    -
    665  // Note: apply the entire column (steps)
    -
    666  const SizeType steps = nrStepsForSweep(j * b, mat_hh.size().cols(), b);
    -
    667  for (SizeType step = 0; step < steps; ++step) {
    -
    668  const SizeType i = j + step;
    -
    669 
    -
    670  const GlobalElementIndex ij_el(i * b, j * b);
    -
    671  const LocalTileIndex ij(dist_hh.localTileIndex(dist_hh.globalTileIndex(ij_el)));
    -
    672 
    -
    673  // Note: reflector with size = 1 must be ignored, except for the last step of the last sweep
    -
    674  // with complex type
    -
    675  const SizeType nrefls = [&]() {
    -
    676  const bool allowSize1 = isComplex_v<T> && j == j_last_sweep && step == steps - 1;
    -
    677  const GlobalElementSize delta(dist_hh.size().rows() - ij_el.row() - 1,
    -
    678  std::min(b, dist_hh.size().cols() - ij_el.col()));
    -
    679  return std::min(b, std::min(delta.rows() - (allowSize1 ? 0 : 1), delta.cols()));
    -
    680  }();
    -
    681 
    -
    682  const TileAccessHelper helper(b, nrefls, dist_hh, dist_e_rt, ij_el);
    -
    683 
    -
    684  if (nrefls < b) {
    -
    685  mat_t.setWidth(nrefls);
    -
    686  mat_v.setWidth(nrefls);
    -
    687  mat_w.setWidth(nrefls);
    -
    688  mat_w2.setHeight(nrefls);
    -
    689  }
    -
    690 
    -
    691  auto [tile_v_unshared, tile_w_unshared] =
    -
    692  helperBackend.computeVW(group_size, ij, helper,
    -
    693  splitTile(mat_hh.read(ij), helper.specHHCompact()), mat_v, mat_t,
    -
    694  mat_w);
    -
    695  auto tile_v = matrix::shareReadWriteTile(ex::make_unique_any_sender(std::move(tile_v_unshared)));
    -
    696  auto tile_w = matrix::shareReadWriteTile(ex::make_unique_any_sender(std::move(tile_w_unshared)));
    -
    697 
    -
    698  for (SizeType j_e = 0; j_e < dist_e_rt.nrTiles().cols(); ++j_e) {
    -
    699  const auto idx_e = helper.topIndexE(j_e);
    -
    700 
    -
    701  if (!helper.affectsMultipleTiles()) {
    -
    702  ex::start_detached(
    -
    703  ex::when_all(ex::just(group_size), tile_v, tile_w,
    -
    704  mat_w2.readwrite(LocalTileIndex(0, j_e)), mat_e_rt.readwrite(idx_e)) |
    -
    705  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    -
    706  dlaf::internal::Policy<B>(thread_priority::normal), ApplyHHToSingleTileRow<B, T>{}));
    -
    707  }
    -
    708  else {
    -
    709  ex::start_detached(
    -
    710  ex::when_all(ex::just(group_size), tile_v, tile_w,
    -
    711  mat_w2.readwrite(LocalTileIndex(0, j_e)), mat_e_rt.readwrite(idx_e),
    -
    712  mat_e_rt.readwrite(helper.bottomIndexE(j_e))) |
    -
    713  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    -
    714  dlaf::internal::Policy<B>(thread_priority::normal), ApplyHHToDoubleTileRow<B, T>{}));
    -
    715  }
    -
    716  }
    -
    717 
    -
    718  mat_t.reset();
    -
    719  mat_v.reset();
    -
    720  mat_w.reset();
    -
    721  mat_w2.reset();
    -
    722  }
    -
    723  }
    -
    724 }
    -
    725 
    -
    726 template <Backend B, Device D, class T>
    -
    727 void BackTransformationT2B<B, D, T>::call(comm::CommunicatorGrid grid, const SizeType band_size,
    -
    728  Matrix<T, D>& mat_e, Matrix<const T, Device::CPU>& mat_hh) {
    -
    729  using pika::execution::thread_priority;
    -
    730  namespace ex = pika::execution::experimental;
    -
    731 
    -
    732  using common::iterate_range2d;
    -
    733  using common::RoundRobin;
    -
    734  using matrix::Panel;
    -
    735  using namespace bt_tridiag;
    -
    736 
    -
    737  if (mat_hh.size().isEmpty() || mat_e.size().isEmpty())
    -
    738  return;
    -
    739 
    -
    740  // Note: if no householder reflectors are going to be applied (in case of trivial matrix)
    -
    741  if (nrSweeps<T>(mat_hh.size().rows()) == 0)
    -
    742  return;
    -
    743 
    -
    744  const SizeType b = band_size;
    -
    745  const SizeType mb = mat_hh.blockSize().rows();
    -
    746  const SizeType group_size = getTuneParameters().bt_band_to_tridiag_hh_apply_group_size;
    -
    747 
    -
    748  const LocalTileSize tiles_per_block(mat_e.blockSize().rows() / b, 1);
    -
    749  matrix::RetiledMatrix<T, D> mat_e_rt(mat_e, tiles_per_block);
    -
    750 
    -
    751  const auto& dist_hh = mat_hh.distribution();
    -
    752  const auto& dist_e_rt = mat_e_rt.distribution();
    -
    753 
    -
    754  // Note: w_tile_sz can store reflectors as they are actually applied, opposed to how they are
    -
    755  // stored in compact form.
    -
    756  //
    -
    757  // e.g. Given b = 4
    -
    758  //
    -
    759  // compact w_tile_sz
    -
    760  // 1 1 1 1 1 0 0 0
    -
    761  // a b c d a 1 0 0
    -
    762  // a b c d a b 1 0
    -
    763  // a b c d a b c 1
    -
    764  // 0 b c d
    -
    765  // 0 0 c d
    -
    766  // 0 0 0 d
    -
    767  const TileElementSize w_tile_sz(2 * b - 1, b);
    -
    768 
    -
    769  const SizeType nlocal_ws =
    -
    770  std::max<SizeType>(1, dist_hh.localNrTiles().rows() * (util::ceilDiv<SizeType>(mb / b, 2) + 1));
    -
    771  const matrix::Distribution dist_ws_hh({nlocal_ws * b, b}, {b, b});
    -
    772  const matrix::Distribution dist_ws_v({nlocal_ws * w_tile_sz.rows(), w_tile_sz.cols()}, w_tile_sz);
    -
    773  const matrix::Distribution dist_ws_w2({nlocal_ws * b, mat_e_rt.size().cols()},
    -
    774  {b, mat_e_rt.blockSize().cols()});
    -
    775 
    -
    776  constexpr std::size_t n_workspaces = 2;
    -
    777 
    -
    778  RoundRobin<Panel<Coord::Col, T, D>> t_panels(n_workspaces, dist_ws_hh);
    -
    779  RoundRobin<Panel<Coord::Col, T, Device::CPU>> hh_panels(n_workspaces, dist_ws_hh);
    -
    780 
    -
    781  RoundRobin<Panel<Coord::Col, T, D>> v_panels(n_workspaces, dist_ws_v);
    -
    782  RoundRobin<Panel<Coord::Col, T, D>> w_panels(n_workspaces, dist_ws_v);
    -
    783 
    -
    784  RoundRobin<Panel<Coord::Row, T, D>> w2_panels(n_workspaces, dist_ws_w2);
    -
    785  RoundRobin<Panel<Coord::Row, T, D>> w2tmp_panels(n_workspaces, dist_ws_w2);
    -
    786 
    -
    787  HHManager<B, D, T> helperBackend(b, n_workspaces, dist_ws_hh, dist_ws_v);
    -
    788 
    -
    789  // Note: This distributed algorithm encompass two communication categories:
    -
    790  // 1. exchange of HH: broadcast + send p2p
    -
    791  // 2. reduction for computing W2: all reduce p2p
    -
    792  // P2P communication can happen out of order since they can be matched via tags, but this is not
    -
    793  // possible for collective operations such as the broadcast.
    -
    794  //
    -
    795  // For this reason, communications of the phase 1 will be ordered with a pipeline. Instead, for the
    -
    796  // second part, with the aim to not over constrain execution of the update, no order will be
    -
    797  // enforced by relying solely on tags.
    -
    798  common::Pipeline<comm::Communicator> mpi_chain_row(grid.rowCommunicator().clone());
    -
    799  common::Pipeline<comm::Communicator> mpi_chain_col(grid.colCommunicator().clone());
    -
    800  const auto mpi_col_comm = ex::just(grid.colCommunicator().clone());
    -
    801 
    -
    802  const SizeType idx_last_sweep_b = (nrSweeps<T>(mat_hh.size().cols()) - 1) / b;
    -
    803  const SizeType maxsteps_b = nrStepsForSweep(0, mat_hh.size().rows(), b);
    -
    804 
    -
    805  // Note: Next two nested `for`s describe a special order loop over the matrix, which allow to
    -
    806  // better schedule communications considering the structure of the algorithm.
    -
    807  //
    -
    808  // Each element depends on:
    -
    809  // - top
    -
    810  // - bottom-right
    -
    811  // - right
    -
    812  //
    -
    813  // This basic rule for dependencies can be described collectively as a mechanism where elements are
    -
    814  // "unlocked" in different epochs, which forms a pattern like if the matrix get scanned not
    -
    815  // perpendicularly to their main axis, but instead it gets scanned by a slightly skewed line that goes
    -
    816  // from top right to bottom left.
    -
    817  //
    -
    818  // 5 x x x x
    -
    819  // 6 4 x x x
    -
    820  // 7 5 3 x x
    -
    821  // 8 6 4 2 x
    -
    822  // 9 7 5 3 1
    -
    823  //
    -
    824  // Elements of the same epoch are somehow "independent" and so they can potentially run in parallel,
    -
    825  // given that previous epoch has been completed. Since scheduling happens sequentially, elements
    -
    826  // of the same epoch will be ordered starting from top-most one, resulting in
    -
    827  //
    -
    828  // 7 x x x x
    -
    829  // 10 5 x x x
    -
    830  // 12 8 3 x x
    -
    831  // 14 11 6 2 x
    -
    832  // 15 13 9 4 1
    -
    833  for (SizeType k = idx_last_sweep_b; k > -maxsteps_b; --k) {
    -
    834  auto& mat_t = t_panels.nextResource();
    -
    835  auto& panel_hh = hh_panels.nextResource();
    -
    836  auto& mat_v = v_panels.nextResource();
    -
    837  auto& mat_w = w_panels.nextResource();
    -
    838  auto& mat_w2 = w2_panels.nextResource();
    -
    839  auto& mat_w2tmp = w2tmp_panels.nextResource();
    -
    840 
    -
    841  for (SizeType i_b = std::abs<SizeType>(k), j_b = std::max<SizeType>(0, k);
    -
    842  i_b < j_b + nrStepsForSweep(j_b * b, mat_hh.size().cols(), b); i_b += 2, ++j_b) {
    -
    843  const SizeType step_b = i_b - j_b;
    -
    844  const GlobalElementIndex ij_el(i_b * b, j_b * b);
    -
    845  const GlobalTileIndex ij_g(dist_hh.globalTileIndex(ij_el));
    -
    846 
    -
    847  const comm::Index2D rank = dist_hh.rankIndex();
    -
    848  const comm::Index2D rankHH = dist_hh.rankGlobalTile(ij_g);
    -
    849 
    -
    850  // Note: reflector with size = 1 must be ignored, except for the last step of the last sweep
    -
    851  // with complex type
    -
    852  const SizeType nrefls = [&]() {
    -
    853  const bool allowSize1 = isComplex_v<T> && j_b == idx_last_sweep_b &&
    -
    854  step_b == nrStepsForSweep(j_b * b, mat_hh.size().cols(), b) - 1;
    -
    855  const GlobalElementSize delta(dist_hh.size().rows() - ij_el.row() - 1,
    -
    856  std::min(b, dist_hh.size().cols() - ij_el.col()));
    -
    857  return std::min(b, std::min(delta.rows() - (allowSize1 ? 0 : 1), delta.cols()));
    -
    858  }();
    -
    859 
    -
    860  const TileAccessHelper helper(b, nrefls, dist_hh, dist_e_rt, ij_el);
    -
    861  const DistIndexing indexing_helper(helper, dist_hh, b, ij_g, i_b);
    -
    862 
    -
    863  if (!indexing_helper.isInvolved())
    -
    864  continue;
    -
    865 
    -
    866  if (nrefls < b) {
    -
    867  mat_t.setWidth(nrefls);
    -
    868  mat_v.setWidth(nrefls);
    -
    869  mat_w.setWidth(nrefls);
    -
    870  mat_w2.setHeight(nrefls);
    -
    871  mat_w2tmp.setHeight(nrefls);
    -
    872  }
    -
    873 
    -
    874  // Note:
    -
    875  // From HH it is possible to extract V that is needed for computing W and W2, both required
    -
    876  // for updating E.
    -
    877 
    -
    878  // Send HH to all involved ranks: broadcast on row + send p2p on col
    -
    879  const LocalTileIndex ij_hh_panel = indexing_helper.wsIndexHH();
    -
    880 
    -
    881  // Broadcast on ROW
    -
    882  if (grid.size().cols() > 1 && rank.row() == rankHH.row()) {
    -
    883  if (rank.col() == rankHH.col()) {
    -
    884  ex::start_detached(comm::scheduleSendBcast(
    -
    885  mpi_chain_row(), splitTile(mat_hh.read(ij_g), helper.specHHCompact())));
    -
    886  }
    -
    887  else {
    -
    888  ex::start_detached(comm::scheduleRecvBcast(mpi_chain_row(), rankHH.col(),
    -
    889  splitTile(panel_hh.readwrite(ij_hh_panel),
    -
    890  helper.specHHCompact(true))));
    -
    891  }
    -
    892  }
    -
    893 
    -
    894  // Jump to the next loop iteration if the local part of E is not affected by the update.
    -
    895  const SizeType ncols_local = dist_e_rt.localNrTiles().cols();
    -
    896  if (ncols_local == 0)
    -
    897  continue;
    -
    898 
    -
    899  // Send P2P on col
    -
    900  if (helper.affectsMultipleRanks()) {
    -
    901  const comm::IndexT_MPI rank_src = rankHH.row();
    -
    902  const comm::IndexT_MPI rank_dst = indexing_helper.rankRowPartner();
    -
    903 
    -
    904  if (rank.row() == rank_src) {
    -
    905  auto tile_hh = rank.col() == rankHH.col()
    -
    906  ? splitTile(mat_hh.read(ij_g), helper.specHHCompact())
    -
    907  : splitTile(panel_hh.read(ij_hh_panel), helper.specHHCompact(true));
    -
    908  ex::start_detached(comm::scheduleSend(mpi_chain_col(), rank_dst, 0, std::move(tile_hh)));
    -
    909  }
    -
    910  else if (rank.row() == rank_dst) {
    -
    911  ex::start_detached(comm::scheduleRecv(mpi_chain_col(), rank_src, 0,
    -
    912  panel_hh.readwrite(ij_hh_panel)));
    -
    913  }
    -
    914  }
    -
    915 
    -
    916  // Batched application increases the communication messages and volume,
    -
    917  // therefore we use it only for local applications.
    -
    918  const SizeType current_group_size = helper.affectsMultipleRanks() ? b : group_size;
    -
    919 
    -
    920  // COMPUTE V and W from HH and T
    -
    921  auto tile_hh = (rankHH == rank)
    -
    922  ? splitTile(mat_hh.read(ij_g), helper.specHHCompact())
    -
    923  : splitTile(panel_hh.read(ij_hh_panel), helper.specHHCompact(true));
    -
    924  auto [tile_v_unshared, tile_w_unshared] =
    -
    925  helperBackend.computeVW(current_group_size, indexing_helper.wsIndexHH(), helper,
    -
    926  std::move(tile_hh), mat_v, mat_t, mat_w);
    -
    927  auto tile_v = matrix::shareReadWriteTile(ex::make_unique_any_sender(std::move(tile_v_unshared)));
    -
    928  auto tile_w = matrix::shareReadWriteTile(ex::make_unique_any_sender(std::move(tile_w_unshared)));
    -
    929 
    -
    930  // UPDATE E
    -
    931  for (SizeType j_e = 0; j_e < ncols_local; ++j_e) {
    -
    932  const SizeType j_e_g = dist_e_rt.template globalTileFromLocalTile<Coord::Col>(j_e);
    -
    933  const LocalTileIndex idx_w2(indexing_helper.wsIndexHH().row(), j_e);
    -
    934 
    -
    935  const GlobalTileIndex idx_e_top = helper.topIndexE(j_e_g);
    -
    936  const auto nb = mat_e_rt.tileSize(idx_e_top).cols();
    -
    937 
    -
    938  // SINGLE ROW UPDATE
    -
    939  if (!helper.affectsMultipleTiles()) {
    -
    940  ex::start_detached(ex::when_all(ex::just(current_group_size), tile_v, tile_w,
    -
    941  splitTile(mat_w2.readwrite(idx_w2), helper.specW2(nb)),
    -
    942  mat_e_rt.readwrite(idx_e_top)) |
    -
    943  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    -
    944  dlaf::internal::Policy<B>(thread_priority::normal),
    -
    945  ApplyHHToSingleTileRow<B, T>{}));
    -
    946  }
    -
    947  // TWO ROWs
    -
    948  else {
    -
    949  const GlobalTileIndex idx_e_bottom = helper.bottomIndexE(j_e_g);
    -
    950 
    -
    951  // TWO ROWs (same RANK)
    -
    952  if (!helper.affectsMultipleRanks()) {
    -
    953  ex::start_detached(
    -
    954  ex::when_all(ex::just(current_group_size), tile_v, tile_w,
    -
    955  splitTile(mat_w2.readwrite(idx_w2), helper.specW2(nb)),
    -
    956  mat_e_rt.readwrite(idx_e_top), mat_e_rt.readwrite(idx_e_bottom)) |
    -
    957  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    -
    958  dlaf::internal::Policy<B>(thread_priority::normal), ApplyHHToDoubleTileRow<B, T>{}));
    -
    959  }
    -
    960  // TWO ROWs TWO RANKs UPDATE (MAIN + PARTNER)
    -
    961  else {
    -
    962  const bool is_top_rank = rank.row() == rankHH.row();
    -
    963  const comm::IndexT_MPI rank_partner =
    -
    964  is_top_rank ? indexing_helper.rankRowPartner() : rankHH.row();
    -
    965 
    -
    966  const comm::IndexT_MPI tag = to_int(j_e + i_b * ncols_local);
    -
    967 
    -
    968  const matrix::SubTileSpec spec_vw = helper.specHH(is_top_rank);
    -
    969 
    -
    970  const auto idx_e = is_top_rank ? idx_e_top : idx_e_bottom;
    -
    971  const auto sz_e = mat_e_rt.tileSize(idx_e);
    -
    972  const matrix::SubTileSpec spec_e{{(is_top_rank ? 1 : 0), 0},
    -
    973  {sz_e.rows() - (is_top_rank ? 1 : 0), sz_e.cols()}};
    -
    974 
    -
    975  auto subtile_v = splitTile(tile_v, spec_vw);
    -
    976  auto subtile_w = splitTile(tile_w, spec_vw);
    -
    977  auto subtile_e_ro = splitTile(mat_e_rt.read(idx_e), spec_e);
    -
    978 
    -
    979  // W2 = V* . E
    -
    980  ex::start_detached(
    -
    981  dlaf::internal::whenAllLift(blas::Op::ConjTrans, blas::Op::NoTrans, T(1),
    -
    982  std::move(subtile_v), std::move(subtile_e_ro), T(0),
    -
    983  splitTile(mat_w2tmp.readwrite(idx_w2), helper.specW2(nb))) |
    -
    984  dlaf::tile::gemm(dlaf::internal::Policy<B>(thread_priority::normal)));
    -
    985 
    -
    986  // Compute final W2 by adding the contribution from the partner rank
    -
    987  ex::start_detached( //
    -
    988  comm::scheduleAllSumP2P<B>(mpi_col_comm, rank_partner, tag,
    -
    989  splitTile(mat_w2tmp.read(idx_w2), helper.specW2(nb)),
    -
    990  splitTile(mat_w2.readwrite(idx_w2), helper.specW2(nb))));
    -
    991 
    -
    992  auto subtile_e = splitTile(mat_e_rt.readwrite(idx_e), spec_e);
    -
    993  // E -= W . W2
    -
    994  ex::start_detached( //
    -
    995  dlaf::internal::whenAllLift(blas::Op::NoTrans, blas::Op::NoTrans, T(-1),
    -
    996  std::move(subtile_w),
    -
    997  splitTile(mat_w2.read(idx_w2), helper.specW2(nb)), T(1),
    -
    998  std::move(subtile_e)) |
    -
    999  dlaf::tile::gemm(dlaf::internal::Policy<B>(thread_priority::normal)));
    -
    1000  }
    -
    1001  }
    -
    1002  }
    -
    1003 
    -
    1004  mat_t.reset();
    -
    1005  panel_hh.reset();
    -
    1006  mat_v.reset();
    -
    1007  mat_w.reset();
    -
    1008  mat_w2tmp.reset();
    -
    1009  mat_w2.reset();
    -
    1010  }
    -
    1011  }
    +
    42 #include <dlaf/matrix/tile.h>
    +
    43 #include <dlaf/sender/policy.h>
    +
    44 #include <dlaf/sender/traits.h>
    +
    45 #include <dlaf/sender/transform.h>
    +
    46 #include <dlaf/sender/when_all_lift.h>
    +
    47 #include <dlaf/tune.h>
    +
    48 #include <dlaf/types.h>
    +
    49 #include <dlaf/util_math.h>
    +
    50 #include <dlaf/util_matrix.h>
    +
    51 
    +
    52 namespace dlaf::eigensolver::internal {
    +
    53 
    +
    54 namespace bt_tridiag {
    +
    55 
    +
    56 template <class T>
    +
    57 matrix::Tile<T, Device::CPU> setupVWellFormed(const SizeType b,
    +
    58  const matrix::Tile<const T, Device::CPU>& tile_hh,
    +
    59  matrix::Tile<T, Device::CPU> tile_v) {
    +
    60  using lapack::lacpy;
    +
    61  using lapack::laset;
    +
    62 
    +
    63  common::internal::SingleThreadedBlasScope single;
    +
    64 
    +
    65  // Note: the size of of tile_hh and tile_v embeds a relevant information about the number of
    +
    66  // reflecotrs and their max size. This will be exploited to correctly setup the well formed
    +
    67  // tile with reflectors in place as they will be applied.
    +
    68  const auto k = tile_v.size().cols();
    +
    69 
    +
    70  // copy from compact representation reflector values (the first component set to 1 is not there)
    +
    71  for (SizeType j = 0; j < k; ++j) {
    +
    72  const auto compact_refl_size =
    +
    73  std::min<SizeType>(tile_v.size().rows() - (1 + j), tile_hh.size().rows() - 1);
    +
    74 
    +
    75  // this is needed because of complex last reflector (i.e. just 1 element long)
    +
    76  if (compact_refl_size == 0)
    +
    77  continue;
    +
    78 
    +
    79  lacpy(blas::Uplo::General, compact_refl_size, 1, tile_hh.ptr({1, j}), tile_hh.ld(),
    +
    80  tile_v.ptr({1 + j, j}), tile_v.ld());
    +
    81  }
    +
    82 
    +
    83  // Note:
    +
    84  // In addition to setting the diagonal to 1 for missing first components, here it zeros out
    +
    85  // both the upper and the lower part. Indeed due to the skewed shape, reflectors do not occupy
    +
    86  // the full tile height, and V should be fully well-formed because the next triangular
    +
    87  // multiplication, i.e. `V . T`, and the gemm `V* . E`, will use V as a general matrix.
    +
    88  laset(blas::Uplo::Upper, tile_v.size().rows(), k, T(0), T(1), tile_v.ptr({0, 0}), tile_v.ld());
    +
    89 
    +
    90  if (tile_v.size().rows() > b)
    +
    91  laset(blas::Uplo::Lower, tile_v.size().rows() - b, k - 1, T(0), T(0), tile_v.ptr({b, 0}),
    +
    92  tile_v.ld());
    +
    93 
    +
    94  return tile_v;
    +
    95 }
    +
    96 
    +
    97 template <class T>
    +
    98 void computeTFactor(const matrix::Tile<const T, Device::CPU>& tile_taus,
    +
    99  const matrix::Tile<const T, Device::CPU>& tile_v,
    +
    100  const matrix::Tile<T, Device::CPU>& tile_t) {
    +
    101  using namespace lapack;
    +
    102 
    +
    103  common::internal::SingleThreadedBlasScope single;
    +
    104 
    +
    105  // taus have to be extracted from the compact form (i.e. first row of the input tile)
    +
    106  std::vector<T> taus;
    +
    107  taus.resize(to_sizet(tile_v.size().cols()));
    +
    108  for (SizeType i = 0; i < to_SizeType(taus.size()); ++i)
    +
    109  taus[to_sizet(i)] = tile_taus({0, i});
    +
    110 
    +
    111  const auto n = tile_v.size().rows();
    +
    112  const auto k = tile_v.size().cols();
    +
    113  larft(Direction::Forward, StoreV::Columnwise, n, k, tile_v.ptr(), tile_v.ld(), taus.data(),
    +
    114  tile_t.ptr(), tile_t.ld());
    +
    115 }
    +
    116 
    +
    117 template <class T>
    +
    118 std::tuple<matrix::Tile<T, Device::CPU>, matrix::Tile<T, Device::CPU>> computeVT(
    +
    119  const SizeType b, const matrix::Tile<const T, Device::CPU>& tile_hh, const SizeType hhr_nb,
    +
    120  matrix::Tile<T, Device::CPU> tile_v, matrix::Tile<T, Device::CPU> tile_t) {
    +
    121  auto tile_v2 = setupVWellFormed(b, tile_hh, std::move(tile_v));
    +
    122  for (SizeType j = 0; j < tile_v2.size().cols(); j += hhr_nb) {
    +
    123  const SizeType jb = std::min(hhr_nb, tile_v2.size().cols() - j);
    +
    124  const SizeType ib = std::min(jb + b - 1, tile_v2.size().rows() - j);
    +
    125  auto subtile_t = tile_t.subTileReference({{j, j}, {jb, jb}});
    +
    126  auto subtile_hh = tile_hh.subTileReference({{0, j}, {1, jb}});
    +
    127  auto subtile_v = tile_v2.subTileReference({{j, j}, {ib, jb}});
    +
    128  computeTFactor(subtile_hh, subtile_v, subtile_t);
    +
    129  }
    +
    130  return std::make_tuple(std::move(tile_v2), std::move(tile_t));
    +
    131 }
    +
    132 
    +
    133 template <class T>
    +
    134 std::tuple<matrix::Tile<T, Device::CPU>, matrix::Tile<T, Device::CPU>> computeVW(
    +
    135  const SizeType b, const matrix::Tile<const T, Device::CPU>& tile_hh, const SizeType hhr_nb,
    +
    136  matrix::Tile<T, Device::CPU> tile_v, matrix::Tile<T, Device::CPU> tile_t,
    +
    137  matrix::Tile<T, Device::CPU> tile_w) {
    +
    138  using namespace blas;
    +
    139 
    +
    140  auto [tile_v2, tile_t2] = computeVT(b, tile_hh, hhr_nb, std::move(tile_v), std::move(tile_t));
    +
    141 
    +
    142  for (SizeType j = 0; j < tile_v2.size().cols(); j += hhr_nb) {
    +
    143  const SizeType jb = std::min(hhr_nb, tile_v2.size().cols() - j);
    +
    144  const SizeType ib = std::min(jb + b - 1, tile_v2.size().rows() - j);
    +
    145  auto subtile_t = tile_t2.subTileReference({{j, j}, {jb, jb}});
    +
    146  auto subtile_v = tile_v2.subTileReference({{j, j}, {ib, jb}});
    +
    147  auto subtile_w = tile_w.subTileReference({{j, j}, {ib, jb}});
    +
    148 
    +
    149  dlaf::tile::internal::trmm3(Side::Right, Uplo::Upper, Op::NoTrans, Diag::NonUnit, T(1), subtile_t,
    +
    150  subtile_v, subtile_w);
    +
    151  }
    +
    152  return std::make_tuple(std::move(tile_v2), std::move(tile_w));
    +
    153 }
    +
    154 
    +
    155 template <class Tile, class CTile>
    +
    156 std::tuple<CTile, CTile, Tile, Tile> applyHHToSingleTileRowSubtileHelper( //
    +
    157  const SizeType j, const SizeType jb, const CTile& tile_v, const CTile& tile_w, const Tile& tile_w2,
    +
    158  const Tile& tile_e) {
    +
    159  DLAF_ASSERT_HEAVY(tile_v.size() == tile_w.size(), tile_v, tile_w);
    +
    160  DLAF_ASSERT_HEAVY(tile_e.size().rows() - 1 == tile_v.size().rows(), tile_e, tile_v);
    +
    161  DLAF_ASSERT_HEAVY(tile_e.size().cols() == tile_w2.size().cols(), tile_e, tile_w2);
    +
    162 
    +
    163  const SizeType ib = tile_v.size().rows() - j;
    +
    164  auto subtile_v = tile_v.subTileReference({{j, j}, {ib, jb}});
    +
    165  auto subtile_w = tile_w.subTileReference({{j, j}, {ib, jb}});
    +
    166  auto subtile_w2 = tile_w2.subTileReference({{0, 0}, {jb, tile_w2.size().cols()}});
    +
    167  auto subtile_e = tile_e.subTileReference({{j + 1, 0}, tile_e.size() - TileElementSize{j + 1, 0}});
    +
    168 
    +
    169  return {std::move(subtile_v), std::move(subtile_w), std::move(subtile_w2), std::move(subtile_e)};
    +
    170 }
    +
    171 
    +
    172 template <Backend B, class T>
    +
    173 struct ApplyHHToSingleTileRow;
    +
    174 
    +
    175 template <class T>
    +
    176 struct ApplyHHToSingleTileRow<Backend::MC, T> {
    +
    177  void operator()(const SizeType hhr_nb, const matrix::Tile<const T, Device::CPU>& tile_v,
    +
    178  const matrix::Tile<const T, Device::CPU>& tile_w,
    +
    179  const matrix::Tile<T, Device::CPU>& tile_w2,
    +
    180  const matrix::Tile<T, Device::CPU>& tile_e) {
    +
    181  using namespace blas;
    +
    182  using tile::internal::gemm;
    +
    183 
    +
    184  common::internal::SingleThreadedBlasScope single;
    +
    185 
    +
    186  for (SizeType j = (util::ceilDiv(tile_v.size().cols(), hhr_nb) - 1) * hhr_nb; j >= 0; j -= hhr_nb) {
    +
    187  const SizeType jb = std::min(hhr_nb, tile_v.size().cols() - j);
    +
    188  auto [subtile_v, subtile_w, subtile_w2, subtile_e] =
    +
    189  applyHHToSingleTileRowSubtileHelper(j, jb, tile_v, tile_w, tile_w2, tile_e);
    +
    190 
    +
    191  // W2 = V* . E
    +
    192  gemm(Op::ConjTrans, Op::NoTrans, T(1), subtile_v, subtile_e, T(0), subtile_w2);
    +
    193  // E -= W . W2
    +
    194  gemm(Op::NoTrans, Op::NoTrans, T(-1), subtile_w, subtile_w2, T(1), subtile_e);
    +
    195  }
    +
    196  }
    +
    197 };
    +
    198 
    +
    199 #ifdef DLAF_WITH_GPU
    +
    200 template <class T>
    +
    201 struct ApplyHHToSingleTileRow<Backend::GPU, T> {
    +
    202  void operator()(cublasHandle_t handle, const SizeType hhr_nb,
    +
    203  const matrix::Tile<const T, Device::GPU>& tile_v,
    +
    204  const matrix::Tile<const T, Device::GPU>& tile_w,
    +
    205  const matrix::Tile<T, Device::GPU>& tile_w2,
    +
    206  const matrix::Tile<T, Device::GPU>& tile_e) {
    +
    207  using namespace blas;
    +
    208  using tile::internal::gemm;
    +
    209 
    +
    210  common::internal::SingleThreadedBlasScope single;
    +
    211 
    +
    212  for (SizeType j = (util::ceilDiv(tile_v.size().cols(), hhr_nb) - 1) * hhr_nb; j >= 0; j -= hhr_nb) {
    +
    213  const SizeType jb = std::min(hhr_nb, tile_v.size().cols() - j);
    +
    214  auto [subtile_v, subtile_w, subtile_w2, subtile_e] =
    +
    215  applyHHToSingleTileRowSubtileHelper(j, jb, tile_v, tile_w, tile_w2, tile_e);
    +
    216 
    +
    217  // W2 = V* . E
    +
    218  gemm(handle, Op::ConjTrans, Op::NoTrans, T(1), subtile_v, subtile_e, T(0), subtile_w2);
    +
    219  // E -= W . W2
    +
    220  gemm(handle, Op::NoTrans, Op::NoTrans, T(-1), subtile_w, subtile_w2, T(1), subtile_e);
    +
    221  }
    +
    222  }
    +
    223 };
    +
    224 #endif
    +
    225 
    +
    226 template <class Tile, class CTile>
    +
    227 std::tuple<CTile, CTile, CTile, CTile, Tile, Tile, Tile> applyHHToDoubleTileRowSubtileHelper( //
    +
    228  const SizeType j, const SizeType jb, const CTile& tile_v, const CTile& tile_w, const Tile& tile_w2,
    +
    229  const Tile& tile_e_top, const Tile& tile_e_bottom) {
    +
    230  DLAF_ASSERT_HEAVY(tile_v.size() == tile_w.size(), tile_v, tile_w);
    +
    231  DLAF_ASSERT_HEAVY(tile_e_top.size().rows() + tile_e_bottom.size().rows() - 1 == tile_v.size().rows(),
    +
    232  tile_e_top, tile_e_bottom, tile_v);
    +
    233  DLAF_ASSERT_HEAVY(tile_e_top.size().cols() == tile_w2.size().cols(), tile_e_top, tile_w2);
    +
    234  DLAF_ASSERT_HEAVY(tile_e_bottom.size().cols() == tile_w2.size().cols(), tile_e_bottom, tile_w2);
    +
    235 
    +
    236  auto subtile_e_top =
    +
    237  tile_e_top.subTileReference({{j + 1, 0}, tile_e_top.size() - TileElementSize{j + 1, 0}});
    +
    238  auto subtile_e_bottom = tile_e_bottom.subTileReference(
    +
    239  {{0, 0},
    +
    240  TileElementSize{std::min(tile_e_bottom.size().rows(), j + jb), tile_e_bottom.size().cols()}});
    +
    241 
    +
    242  matrix::SubTileSpec spec_top{{j, j}, {subtile_e_top.size().rows(), jb}};
    +
    243  matrix::SubTileSpec spec_bottom{{tile_e_top.size().rows() - 1, j},
    +
    244  {subtile_e_bottom.size().rows(), jb}};
    +
    245 
    +
    246  auto subtile_v_top = tile_v.subTileReference(spec_top);
    +
    247  auto subtile_v_bottom = tile_v.subTileReference(spec_bottom);
    +
    248  auto subtile_w_top = tile_w.subTileReference(spec_top);
    +
    249  auto subtile_w_bottom = tile_w.subTileReference(spec_bottom);
    +
    250  auto subtile_w2 = tile_w2.subTileReference({{0, 0}, {jb, tile_w2.size().cols()}});
    +
    251 
    +
    252  return {std::move(subtile_v_top), std::move(subtile_v_bottom), std::move(subtile_w_top),
    +
    253  std::move(subtile_w_bottom), std::move(subtile_w2), std::move(subtile_e_top),
    +
    254  std::move(subtile_e_bottom)};
    +
    255 }
    +
    256 
    +
    257 template <Backend B, class T>
    +
    258 struct ApplyHHToDoubleTileRow;
    +
    259 
    +
    260 template <class T>
    +
    261 struct ApplyHHToDoubleTileRow<Backend::MC, T> {
    +
    262  void operator()(const SizeType hhr_nb, const matrix::Tile<const T, Device::CPU>& tile_v,
    +
    263  const matrix::Tile<const T, Device::CPU>& tile_w,
    +
    264  const matrix::Tile<T, Device::CPU>& tile_w2,
    +
    265  const matrix::Tile<T, Device::CPU>& tile_e_top,
    +
    266  const matrix::Tile<T, Device::CPU>& tile_e_bottom) {
    +
    267  using namespace blas;
    +
    268  using tile::internal::gemm;
    +
    269 
    +
    270  common::internal::SingleThreadedBlasScope single;
    +
    271 
    +
    272  for (SizeType j = (util::ceilDiv(tile_v.size().cols(), hhr_nb) - 1) * hhr_nb; j >= 0; j -= hhr_nb) {
    +
    273  const SizeType jb = std::min(hhr_nb, tile_v.size().cols() - j);
    +
    274  auto [subtile_v_top, subtile_v_bottom, subtile_w_top, subtile_w_bottom, subtile_w2, subtile_e_top,
    +
    275  subtile_e_bottom] =
    +
    276  applyHHToDoubleTileRowSubtileHelper(j, jb, tile_v, tile_w, tile_w2, tile_e_top, tile_e_bottom);
    +
    277 
    +
    278  // W2 = V* . E
    +
    279  gemm(Op::ConjTrans, Op::NoTrans, T(1), subtile_v_top, subtile_e_top, T(0), subtile_w2);
    +
    280  gemm(Op::ConjTrans, Op::NoTrans, T(1), subtile_v_bottom, subtile_e_bottom, T(1), subtile_w2);
    +
    281  // E -= W . W2
    +
    282  gemm(Op::NoTrans, Op::NoTrans, T(-1), subtile_w_top, subtile_w2, T(1), subtile_e_top);
    +
    283  gemm(Op::NoTrans, Op::NoTrans, T(-1), subtile_w_bottom, subtile_w2, T(1), subtile_e_bottom);
    +
    284  }
    +
    285  }
    +
    286 };
    +
    287 
    +
    288 #ifdef DLAF_WITH_GPU
    +
    289 template <class T>
    +
    290 struct ApplyHHToDoubleTileRow<Backend::GPU, T> {
    +
    291  void operator()(cublasHandle_t handle, const SizeType hhr_nb,
    +
    292  const matrix::Tile<const T, Device::GPU>& tile_v,
    +
    293  const matrix::Tile<const T, Device::GPU>& tile_w,
    +
    294  const matrix::Tile<T, Device::GPU>& tile_w2,
    +
    295  const matrix::Tile<T, Device::GPU>& tile_e_top,
    +
    296  const matrix::Tile<T, Device::GPU>& tile_e_bottom) {
    +
    297  using namespace blas;
    +
    298  using tile::internal::gemm;
    +
    299 
    +
    300  common::internal::SingleThreadedBlasScope single;
    +
    301 
    +
    302  for (SizeType j = (util::ceilDiv(tile_v.size().cols(), hhr_nb) - 1) * hhr_nb; j >= 0; j -= hhr_nb) {
    +
    303  const SizeType jb = std::min(hhr_nb, tile_v.size().cols() - j);
    +
    304  auto [subtile_v_top, subtile_v_bottom, subtile_w_top, subtile_w_bottom, subtile_w2, subtile_e_top,
    +
    305  subtile_e_bottom] =
    +
    306  applyHHToDoubleTileRowSubtileHelper(j, jb, tile_v, tile_w, tile_w2, tile_e_top, tile_e_bottom);
    +
    307 
    +
    308  // W2 = V* . E
    +
    309  gemm(handle, Op::ConjTrans, Op::NoTrans, T(1), subtile_v_top, subtile_e_top, T(0), subtile_w2);
    +
    310  gemm(handle, Op::ConjTrans, Op::NoTrans, T(1), subtile_v_bottom, subtile_e_bottom, T(1),
    +
    311  subtile_w2);
    +
    312  // E -= W . W2
    +
    313  gemm(handle, Op::NoTrans, Op::NoTrans, T(-1), subtile_w_top, subtile_w2, T(1), subtile_e_top);
    +
    314  gemm(handle, Op::NoTrans, Op::NoTrans, T(-1), subtile_w_bottom, subtile_w2, T(1),
    +
    315  subtile_e_bottom);
    +
    316  }
    +
    317  }
    +
    318 };
    +
    319 #endif
    +
    320 
    +
    321 struct TileAccessHelper {
    +
    322  TileAccessHelper(const SizeType b, const SizeType nrefls, const matrix::Distribution& dist_hh,
    +
    323  const matrix::Distribution& dist_e, const GlobalElementIndex offset)
    +
    324  : nrefls_(nrefls), input_spec_{dist_hh.tileElementIndex(offset),
    +
    325  {std::min(b, dist_hh.size().rows() - offset.row()),
    +
    326  std::min(b, dist_hh.size().cols() - offset.col())}},
    +
    327  index_e_(dist_e.globalTileFromGlobalElement<Coord::Row>(offset.row())) {
    +
    328  // Note:
    +
    329  // Next logic is about detecting the available application space for reflectors, which once
    +
    330  // extracted they expands to a matrix with 2 * b - 1 height.
    +
    331  //
    +
    332  // There are two main scenarios:
    +
    333  // - reflectors involves rows of a single tile;
    +
    334  // - reflectors involves rows across two different tiles.
    +
    335 
    +
    336  rows_v_ = std::min(2 * b, dist_e.size().rows() - offset.row()) - 1;
    +
    337  rows_v_top_ = std::min(rows_v_, b - 1);
    +
    338 
    +
    339  // Note:
    +
    340  // In general, keep in mind that the first row of the top tile_e should not be touched.
    +
    341  // The structs ApplyHHTo... take care of it.
    +
    342  //
    +
    343  // e.g. b = 4
    +
    344  // reflectors matrix
    +
    345  // X X X X
    +
    346  // 1 0 0 0 X X X X
    +
    347  // a 1 0 0 X X X X
    +
    348  // a b 1 0 X X X X
    +
    349  // -------
    +
    350  // a b c 1 Y Y Y Y
    +
    351  // 0 b c d Y Y Y Y
    +
    352  // 0 0 c d Y Y Y Y
    +
    353  // 0 0 0 d Y Y Y Y
    +
    354  //
    +
    355  // From the drawing above, it is possible to see the dashed tile separation between X and Y,
    +
    356  // and how the reflectors on the left are going to be applied. In particular, the first row of
    +
    357  // the upper tile is not affected.
    +
    358 
    +
    359  // Note:
    +
    360  // A single tile is involved only if it is the last row tile, so by construction reflectors will be
    +
    361  // applied to a single tile;
    +
    362  across_tiles_ = !(index_e_ == dist_e.nrTiles().rows() - 1);
    +
    363 
    +
    364  across_ranks_ = across_tiles_ && (dist_e.rankGlobalTile<Coord::Row>(index_e_) !=
    +
    365  dist_e.rankGlobalTile<Coord::Row>(index_e_ + 1));
    +
    366  }
    +
    367 
    +
    368  // Return true if the application of Householder reflectors involves multiple tiles
    +
    369  bool affectsMultipleTiles() const noexcept {
    +
    370  return across_tiles_;
    +
    371  }
    +
    372 
    +
    373  // Return true if the application of Householder reflectors involves multiple ranks
    +
    374  bool affectsMultipleRanks() const noexcept {
    +
    375  return across_ranks_;
    +
    376  }
    +
    377 
    +
    378  // Return SubTileSpec to use for accessing Householder reflectors in compact form
    +
    379  //
    +
    380  // SubTileSpec returned points to the sub-block in the full tile containing the HH data in compact
    +
    381  // form. If @p reset_origin is true, then the origin component of the SubTileSpec is resetted and
    +
    382  // it will just describe the size of the sub-block containing the Householder reflectors (useful
    +
    383  // for panel access which might not have full-tiles).
    +
    384  matrix::SubTileSpec specHHCompact(const bool reset_origin = false) const noexcept {
    +
    385  if (reset_origin)
    +
    386  return {{0, 0}, input_spec_.size};
    +
    387  return input_spec_;
    +
    388  }
    +
    389 
    +
    390  // Return SubTileSpec to use for accessing Householder reflectors in well formed form
    +
    391  matrix::SubTileSpec specHH() const noexcept {
    +
    392  return {{0, 0}, {rows_v_, nrefls_}};
    +
    393  }
    +
    394 
    +
    395  // Return SubTileSpec to use for accessing Householder reflectors in well formed form
    +
    396  // for the distributed case.
    +
    397  matrix::SubTileSpec specHH(bool is_top_rank) const noexcept {
    +
    398  if (is_top_rank)
    +
    399  return {{0, 0}, {rows_v_top_, nrefls_}};
    +
    400  DLAF_ASSERT_MODERATE(affectsMultipleTiles(), affectsMultipleTiles());
    +
    401  return {{rows_v_top_, 0}, {rows_v_ - rows_v_top_, nrefls_}};
    +
    402  }
    +
    403 
    +
    404  // Return SubTileSpec to use for accessing T factor
    +
    405  matrix::SubTileSpec specT() const noexcept {
    +
    406  return {{0, 0}, {nrefls_, nrefls_}};
    +
    407  }
    +
    408 
    +
    409  matrix::SubTileSpec specW2(const SizeType cols) const noexcept {
    +
    410  return {{0, 0}, {nrefls_, cols}};
    +
    411  }
    +
    412 
    +
    413  GlobalTileIndex topIndexE(const SizeType j) const noexcept {
    +
    414  return {index_e_, j};
    +
    415  }
    +
    416 
    +
    417  GlobalTileIndex bottomIndexE(const SizeType j) const noexcept {
    +
    418  DLAF_ASSERT_MODERATE(affectsMultipleTiles(), affectsMultipleTiles());
    +
    419  return {index_e_ + 1, j};
    +
    420  }
    +
    421 
    +
    422 private:
    +
    423  SizeType nrefls_;
    +
    424  matrix::SubTileSpec input_spec_;
    +
    425  SizeType index_e_;
    +
    426  SizeType rows_v_;
    +
    427  SizeType rows_v_top_;
    +
    428 
    +
    429  bool across_tiles_;
    +
    430  bool across_ranks_;
    +
    431 };
    +
    432 
    +
    445 struct DistIndexing {
    +
    446  DistIndexing(const TileAccessHelper& helper, const matrix::Distribution& dist_hh, const SizeType b,
    +
    447  const GlobalTileIndex& ij, const SizeType& ij_b_row)
    +
    448  : dist_hh(dist_hh), b(b), mb(dist_hh.blockSize().rows()), helper(helper), ij(ij),
    +
    449  ij_b_row(ij_b_row) {
    +
    450  rank = dist_hh.rankIndex();
    +
    451  rankHH = dist_hh.rankGlobalTile(ij);
    +
    452  n_ws_per_block = to_SizeType(static_cast<size_t>(std::ceil(mb / b / 2.0f)) + 1);
    +
    453  }
    +
    454 
    +
    455  comm::IndexT_MPI rankRowPartner() const {
    +
    456  return (rankHH.row() + 1) % dist_hh.commGridSize().rows();
    +
    457  }
    +
    458 
    +
    459  bool isInvolved() const {
    +
    460  const bool isSameRow = rank.row() == rankHH.row();
    +
    461  const bool isPartnerRow = rank.row() == rankRowPartner();
    +
    462  return isSameRow || (isPartnerRow && helper.affectsMultipleRanks());
    +
    463  }
    +
    464 
    +
    465  LocalTileIndex wsIndexHH() const {
    +
    466  const SizeType row = [&]() -> SizeType {
    +
    467  if (rank.row() == rankHH.row()) {
    +
    468  // Note: index starts at 1 (0 is the extra workspace), moreover max half blocks will run in parallel
    +
    469  const SizeType intra_idx = 1 + (ij_b_row % (mb / b)) / 2;
    +
    470  DLAF_ASSERT_HEAVY(intra_idx < n_ws_per_block, intra_idx, n_ws_per_block);
    +
    471  return dist_hh.localTileFromGlobalTile<Coord::Row>(ij.row()) * n_ws_per_block + intra_idx;
    +
    472  }
    +
    473  else {
    +
    474  DLAF_ASSERT_HEAVY(helper.affectsMultipleRanks() && (rank.row() == rankRowPartner()),
    +
    475  helper.affectsMultipleRanks(), rank.row(), rankRowPartner());
    +
    476  return dist_hh.localNrTiles().isEmpty()
    +
    477  ? 0
    +
    478  : dist_hh.localTileFromGlobalTile<Coord::Row>(ij.row() + 1) * n_ws_per_block;
    +
    479  }
    +
    480  }();
    +
    481  return {row, 0};
    +
    482  }
    +
    483 
    +
    484 protected:
    +
    485  matrix::Distribution dist_hh;
    +
    486  SizeType b;
    +
    487  SizeType mb;
    +
    488  SizeType n_ws_per_block;
    +
    489 
    +
    490  TileAccessHelper helper;
    +
    491 
    +
    492  comm::Index2D rank;
    +
    493  comm::Index2D rankHH;
    +
    494 
    +
    495  GlobalTileIndex ij;
    +
    496  SizeType ij_b_row;
    +
    497 };
    +
    498 
    +
    499 template <Backend B, Device D, class T>
    +
    500 struct HHManager;
    +
    501 
    +
    502 template <class T>
    +
    503 struct HHManager<Backend::MC, Device::CPU, T> {
    +
    504  static constexpr Backend B = Backend::MC;
    +
    505  static constexpr Device D = Device::CPU;
    +
    506 
    +
    507  HHManager(const SizeType b, const std::size_t, matrix::Distribution, matrix::Distribution) : b(b) {}
    +
    508 
    +
    509  template <class SenderHH>
    +
    510  auto computeVW(const SizeType nb_apply, const LocalTileIndex ij, const TileAccessHelper& helper,
    +
    511  SenderHH&& tile_hh, matrix::Panel<Coord::Col, T, D>& mat_v,
    +
    512  matrix::Panel<Coord::Col, T, D>& mat_t, matrix::Panel<Coord::Col, T, D>& mat_w) {
    +
    513  namespace ex = pika::execution::experimental;
    +
    514 
    +
    515  return dlaf::internal::whenAllLift(b, std::forward<SenderHH>(tile_hh), nb_apply,
    +
    516  splitTile(mat_v.readwrite(ij), helper.specHH()),
    +
    517  splitTile(mat_t.readwrite(ij), helper.specT()),
    +
    518  splitTile(mat_w.readwrite(ij), helper.specHH())) |
    +
    519  dlaf::internal::transform(dlaf::internal::Policy<Backend::MC>(), bt_tridiag::computeVW<T>) |
    +
    520  ex::split_tuple();
    +
    521  }
    +
    522 
    +
    523 protected:
    +
    524  const SizeType b;
    +
    525 };
    +
    526 
    +
    527 #ifdef DLAF_WITH_GPU
    +
    528 template <class T>
    +
    529 struct HHManager<Backend::GPU, Device::GPU, T> {
    +
    530  static constexpr Backend B = Backend::GPU;
    +
    531  static constexpr Device D = Device::GPU;
    +
    532 
    +
    533  HHManager(const SizeType b, const std::size_t n_workspaces, matrix::Distribution dist_t,
    +
    534  matrix::Distribution dist_w)
    +
    535  : b(b), t_panels_h(n_workspaces, dist_t), w_panels_h(n_workspaces, dist_w) {}
    +
    536 
    +
    537  template <class SenderHH>
    +
    538  auto computeVW(const SizeType hhr_nb, const LocalTileIndex ij, const TileAccessHelper& helper,
    +
    539  SenderHH&& tile_hh, matrix::Panel<Coord::Col, T, D>& mat_v,
    +
    540  matrix::Panel<Coord::Col, T, D>& mat_t, matrix::Panel<Coord::Col, T, D>& mat_w) {
    +
    541  namespace ex = pika::execution::experimental;
    +
    542 
    +
    543  auto& mat_v_h = w_panels_h.nextResource();
    +
    544  auto& mat_t_h = t_panels_h.nextResource();
    +
    545 
    +
    546  const LocalTileIndex ij_t = ij;
    +
    547  const matrix::SubTileSpec t_spec = helper.specT();
    +
    548 
    +
    549  auto [tile_v_h, tile_t_h] =
    +
    550  dlaf::internal::whenAllLift(b, std::forward<SenderHH>(tile_hh), hhr_nb,
    +
    551  splitTile(mat_v_h.readwrite(ij), helper.specHH()),
    +
    552  splitTile(mat_t_h.readwrite(ij_t), t_spec)) |
    +
    553  dlaf::internal::transform(dlaf::internal::Policy<Backend::MC>(), computeVT<T>) |
    +
    554  ex::split_tuple();
    +
    555 
    +
    556  auto copyVTandComputeW =
    +
    557  [b = this->b, hhr_nb](cublasHandle_t handle, const matrix::Tile<const T, Device::CPU>& tile_v_h,
    +
    558  const matrix::Tile<const T, Device::CPU>& tile_t_h,
    +
    559  matrix::Tile<T, Device::GPU>& tile_v, matrix::Tile<T, Device::GPU>& tile_t,
    +
    560  matrix::Tile<T, Device::GPU>& tile_w) {
    +
    561  whip::stream_t stream;
    +
    562  DLAF_GPUBLAS_CHECK_ERROR(cublasGetStream(handle, &stream));
    +
    563 
    +
    564  matrix::internal::copy(tile_v_h, tile_v, stream);
    +
    565  matrix::internal::copy(tile_t_h, tile_t, stream);
    +
    566 
    +
    567  // W = V . T
    +
    568  using namespace blas;
    +
    569  for (SizeType j = 0; j < tile_v_h.size().cols(); j += hhr_nb) {
    +
    570  const SizeType jb = std::min(hhr_nb, tile_v_h.size().cols() - j);
    +
    571  const SizeType ib = std::min(jb + b - 1, tile_v_h.size().rows() - j);
    +
    572  auto subtile_t = tile_t.subTileReference({{j, j}, {jb, jb}});
    +
    573  auto subtile_v = tile_v.subTileReference({{j, j}, {ib, jb}});
    +
    574  auto subtile_w = tile_w.subTileReference({{j, j}, {ib, jb}});
    +
    575 
    +
    576  dlaf::tile::internal::trmm3(handle, Side::Right, Uplo::Upper, Op::NoTrans, Diag::NonUnit,
    +
    577  T(1), subtile_t, subtile_v, subtile_w);
    +
    578  }
    +
    579 
    +
    580  return std::make_tuple(std::move(tile_v), std::move(tile_w));
    +
    581  };
    +
    582 
    +
    583  return ex::when_all(std::move(tile_v_h), std::move(tile_t_h),
    +
    584  splitTile(mat_v.readwrite(ij), helper.specHH()),
    +
    585  splitTile(mat_t.readwrite(ij_t), t_spec),
    +
    586  splitTile(mat_w.readwrite(ij), helper.specHH())) |
    +
    587  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    +
    588  dlaf::internal::Policy<Backend::GPU>(), copyVTandComputeW) |
    +
    589  ex::split_tuple();
    +
    590  }
    +
    591 
    +
    592 protected:
    +
    593  const SizeType b;
    +
    594  common::RoundRobin<matrix::Panel<Coord::Col, T, Device::CPU>> t_panels_h;
    +
    595  common::RoundRobin<matrix::Panel<Coord::Col, T, Device::CPU>> w_panels_h;
    +
    596 };
    +
    597 #endif
    +
    598 }
    +
    599 
    +
    600 template <Backend B, Device D, class T>
    +
    601 void BackTransformationT2B<B, D, T>::call(const SizeType band_size, Matrix<T, D>& mat_e,
    +
    602  Matrix<const T, Device::CPU>& mat_hh) {
    +
    603  using pika::execution::thread_priority;
    +
    604  namespace ex = pika::execution::experimental;
    +
    605 
    +
    606  using common::iterate_range2d;
    +
    607  using common::RoundRobin;
    +
    608  using matrix::Panel;
    +
    609  using namespace bt_tridiag;
    +
    610 
    +
    611  if (mat_hh.size().isEmpty() || mat_e.size().isEmpty())
    +
    612  return;
    +
    613 
    +
    614  // Note: if no householder reflectors are going to be applied (in case of trivial matrix)
    +
    615  if (mat_hh.size().rows() <= (dlaf::isComplex_v<T> ? 1 : 2))
    +
    616  return;
    +
    617 
    +
    618  const SizeType b = band_size;
    +
    619  const SizeType group_size = getTuneParameters().bt_band_to_tridiag_hh_apply_group_size;
    +
    620  const SizeType nsweeps = nrSweeps<T>(mat_hh.size().cols());
    +
    621 
    +
    622  const LocalTileSize tiles_per_block(mat_e.blockSize().rows() / b, 1);
    +
    623  Matrix<T, D> mat_e_rt = mat_e.retiledSubPipeline(tiles_per_block);
    +
    624 
    +
    625  const auto& dist_hh = mat_hh.distribution();
    +
    626  const auto& dist_e_rt = mat_e_rt.distribution();
    +
    627 
    +
    628  // Note: w_tile_sz can store reflectors as they are actually applied, opposed to how they are
    +
    629  // stored in compact form.
    +
    630  //
    +
    631  // e.g. Given b = 4
    +
    632  //
    +
    633  // compact w_tile_sz
    +
    634  // 1 1 1 1 1 0 0 0
    +
    635  // a b c d a 1 0 0
    +
    636  // a b c d a b 1 0
    +
    637  // a b c d a b c 1
    +
    638  // 0 b c d
    +
    639  // 0 0 c d
    +
    640  // 0 0 0 d
    +
    641  const TileElementSize w_tile_sz(2 * b - 1, b);
    +
    642 
    +
    643  const SizeType dist_w_rows = mat_e_rt.nrTiles().rows() * w_tile_sz.rows();
    +
    644  const matrix::Distribution dist_w({dist_w_rows, b}, w_tile_sz);
    +
    645  const matrix::Distribution dist_t({mat_hh.size().rows(), b}, {b, b});
    +
    646  const matrix::Distribution dist_w2({b, mat_e_rt.size().cols()}, {b, mat_e_rt.blockSize().cols()});
    +
    647 
    +
    648  constexpr std::size_t n_workspaces = 2;
    +
    649  RoundRobin<Panel<Coord::Col, T, D>> t_panels(n_workspaces, dist_t);
    +
    650  RoundRobin<Panel<Coord::Col, T, D>> v_panels(n_workspaces, dist_w);
    +
    651  RoundRobin<Panel<Coord::Col, T, D>> w_panels(n_workspaces, dist_w);
    +
    652  RoundRobin<Panel<Coord::Row, T, D>> w2_panels(n_workspaces, dist_w2);
    +
    653 
    +
    654  HHManager<B, D, T> helperBackend(b, n_workspaces, dist_t, dist_w);
    +
    655 
    +
    656  // Note: sweep are on diagonals, steps are on verticals
    +
    657  const SizeType j_last_sweep = (nsweeps - 1) / b;
    +
    658  for (SizeType j = j_last_sweep; j >= 0; --j) {
    +
    659  auto& mat_t = t_panels.nextResource();
    +
    660  auto& mat_v = v_panels.nextResource();
    +
    661  auto& mat_w = w_panels.nextResource();
    +
    662  auto& mat_w2 = w2_panels.nextResource();
    +
    663 
    +
    664  // Note: apply the entire column (steps)
    +
    665  const SizeType steps = nrStepsForSweep(j * b, mat_hh.size().cols(), b);
    +
    666  for (SizeType step = 0; step < steps; ++step) {
    +
    667  const SizeType i = j + step;
    +
    668 
    +
    669  const GlobalElementIndex ij_el(i * b, j * b);
    +
    670  const LocalTileIndex ij(dist_hh.localTileIndex(dist_hh.globalTileIndex(ij_el)));
    +
    671 
    +
    672  // Note: reflector with size = 1 must be ignored, except for the last step of the last sweep
    +
    673  // with complex type
    +
    674  const SizeType nrefls = [&]() {
    +
    675  const bool allowSize1 = isComplex_v<T> && j == j_last_sweep && step == steps - 1;
    +
    676  const GlobalElementSize delta(dist_hh.size().rows() - ij_el.row() - 1,
    +
    677  std::min(b, dist_hh.size().cols() - ij_el.col()));
    +
    678  return std::min(b, std::min(delta.rows() - (allowSize1 ? 0 : 1), delta.cols()));
    +
    679  }();
    +
    680 
    +
    681  const TileAccessHelper helper(b, nrefls, dist_hh, dist_e_rt, ij_el);
    +
    682 
    +
    683  if (nrefls < b) {
    +
    684  mat_t.setWidth(nrefls);
    +
    685  mat_v.setWidth(nrefls);
    +
    686  mat_w.setWidth(nrefls);
    +
    687  mat_w2.setHeight(nrefls);
    +
    688  }
    +
    689 
    +
    690  auto [tile_v_unshared, tile_w_unshared] =
    +
    691  helperBackend.computeVW(group_size, ij, helper,
    +
    692  splitTile(mat_hh.read(ij), helper.specHHCompact()), mat_v, mat_t,
    +
    693  mat_w);
    +
    694  auto tile_v = matrix::shareReadWriteTile(ex::make_unique_any_sender(std::move(tile_v_unshared)));
    +
    695  auto tile_w = matrix::shareReadWriteTile(ex::make_unique_any_sender(std::move(tile_w_unshared)));
    +
    696 
    +
    697  for (SizeType j_e = 0; j_e < dist_e_rt.nrTiles().cols(); ++j_e) {
    +
    698  const auto idx_e = helper.topIndexE(j_e);
    +
    699 
    +
    700  if (!helper.affectsMultipleTiles()) {
    +
    701  ex::start_detached(
    +
    702  ex::when_all(ex::just(group_size), tile_v, tile_w,
    +
    703  mat_w2.readwrite(LocalTileIndex(0, j_e)), mat_e_rt.readwrite(idx_e)) |
    +
    704  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    +
    705  dlaf::internal::Policy<B>(thread_priority::normal), ApplyHHToSingleTileRow<B, T>{}));
    +
    706  }
    +
    707  else {
    +
    708  ex::start_detached(
    +
    709  ex::when_all(ex::just(group_size), tile_v, tile_w,
    +
    710  mat_w2.readwrite(LocalTileIndex(0, j_e)), mat_e_rt.readwrite(idx_e),
    +
    711  mat_e_rt.readwrite(helper.bottomIndexE(j_e))) |
    +
    712  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    +
    713  dlaf::internal::Policy<B>(thread_priority::normal), ApplyHHToDoubleTileRow<B, T>{}));
    +
    714  }
    +
    715  }
    +
    716 
    +
    717  mat_t.reset();
    +
    718  mat_v.reset();
    +
    719  mat_w.reset();
    +
    720  mat_w2.reset();
    +
    721  }
    +
    722  }
    +
    723 }
    +
    724 
    +
    725 template <Backend B, Device D, class T>
    +
    726 void BackTransformationT2B<B, D, T>::call(comm::CommunicatorGrid grid, const SizeType band_size,
    +
    727  Matrix<T, D>& mat_e, Matrix<const T, Device::CPU>& mat_hh) {
    +
    728  using pika::execution::thread_priority;
    +
    729  namespace ex = pika::execution::experimental;
    +
    730 
    +
    731  using common::iterate_range2d;
    +
    732  using common::RoundRobin;
    +
    733  using matrix::Panel;
    +
    734  using namespace bt_tridiag;
    +
    735 
    +
    736  if (mat_hh.size().isEmpty() || mat_e.size().isEmpty())
    +
    737  return;
    +
    738 
    +
    739  // Note: if no householder reflectors are going to be applied (in case of trivial matrix)
    +
    740  if (nrSweeps<T>(mat_hh.size().rows()) == 0)
    +
    741  return;
    +
    742 
    +
    743  const SizeType b = band_size;
    +
    744  const SizeType mb = mat_hh.blockSize().rows();
    +
    745  const SizeType group_size = getTuneParameters().bt_band_to_tridiag_hh_apply_group_size;
    +
    746 
    +
    747  const LocalTileSize tiles_per_block(mat_e.blockSize().rows() / b, 1);
    +
    748  Matrix<T, D> mat_e_rt = mat_e.retiledSubPipeline(tiles_per_block);
    +
    749 
    +
    750  const auto& dist_hh = mat_hh.distribution();
    +
    751  const auto& dist_e_rt = mat_e_rt.distribution();
    +
    752 
    +
    753  // Note: w_tile_sz can store reflectors as they are actually applied, opposed to how they are
    +
    754  // stored in compact form.
    +
    755  //
    +
    756  // e.g. Given b = 4
    +
    757  //
    +
    758  // compact w_tile_sz
    +
    759  // 1 1 1 1 1 0 0 0
    +
    760  // a b c d a 1 0 0
    +
    761  // a b c d a b 1 0
    +
    762  // a b c d a b c 1
    +
    763  // 0 b c d
    +
    764  // 0 0 c d
    +
    765  // 0 0 0 d
    +
    766  const TileElementSize w_tile_sz(2 * b - 1, b);
    +
    767 
    +
    768  const SizeType nlocal_ws =
    +
    769  std::max<SizeType>(1, dist_hh.localNrTiles().rows() * (util::ceilDiv<SizeType>(mb / b, 2) + 1));
    +
    770  const matrix::Distribution dist_ws_hh({nlocal_ws * b, b}, {b, b});
    +
    771  const matrix::Distribution dist_ws_v({nlocal_ws * w_tile_sz.rows(), w_tile_sz.cols()}, w_tile_sz);
    +
    772  const matrix::Distribution dist_ws_w2({nlocal_ws * b, mat_e_rt.size().cols()},
    +
    773  {b, mat_e_rt.blockSize().cols()});
    +
    774 
    +
    775  constexpr std::size_t n_workspaces = 2;
    +
    776 
    +
    777  RoundRobin<Panel<Coord::Col, T, D>> t_panels(n_workspaces, dist_ws_hh);
    +
    778  RoundRobin<Panel<Coord::Col, T, Device::CPU>> hh_panels(n_workspaces, dist_ws_hh);
    +
    779 
    +
    780  RoundRobin<Panel<Coord::Col, T, D>> v_panels(n_workspaces, dist_ws_v);
    +
    781  RoundRobin<Panel<Coord::Col, T, D>> w_panels(n_workspaces, dist_ws_v);
    +
    782 
    +
    783  RoundRobin<Panel<Coord::Row, T, D>> w2_panels(n_workspaces, dist_ws_w2);
    +
    784  RoundRobin<Panel<Coord::Row, T, D>> w2tmp_panels(n_workspaces, dist_ws_w2);
    +
    785 
    +
    786  HHManager<B, D, T> helperBackend(b, n_workspaces, dist_ws_hh, dist_ws_v);
    +
    787 
    +
    788  // Note: This distributed algorithm encompass two communication categories:
    +
    789  // 1. exchange of HH: broadcast + send p2p
    +
    790  // 2. reduction for computing W2: all reduce p2p
    +
    791  // P2P communication can happen out of order since they can be matched via tags, but this is not
    +
    792  // possible for collective operations such as the broadcast.
    +
    793  //
    +
    794  // For this reason, communications of the phase 1 will be ordered with a pipeline. Instead, for the
    +
    795  // second part, with the aim to not over constrain execution of the update, no order will be
    +
    796  // enforced by relying solely on tags.
    +
    797  common::Pipeline<comm::Communicator> mpi_chain_row(grid.rowCommunicator().clone());
    +
    798  common::Pipeline<comm::Communicator> mpi_chain_col(grid.colCommunicator().clone());
    +
    799  const auto mpi_col_comm = ex::just(grid.colCommunicator().clone());
    +
    800 
    +
    801  const SizeType idx_last_sweep_b = (nrSweeps<T>(mat_hh.size().cols()) - 1) / b;
    +
    802  const SizeType maxsteps_b = nrStepsForSweep(0, mat_hh.size().rows(), b);
    +
    803 
    +
    804  // Note: Next two nested `for`s describe a special order loop over the matrix, which allow to
    +
    805  // better schedule communications considering the structure of the algorithm.
    +
    806  //
    +
    807  // Each element depends on:
    +
    808  // - top
    +
    809  // - bottom-right
    +
    810  // - right
    +
    811  //
    +
    812  // This basic rule for dependencies can be described collectively as a mechanism where elements are
    +
    813  // "unlocked" in different epochs, which forms a pattern like if the matrix get scanned not
    +
    814  // perpendicularly to their main axis, but instead it gets scanned by a slightly skewed line that goes
    +
    815  // from top right to bottom left.
    +
    816  //
    +
    817  // 5 x x x x
    +
    818  // 6 4 x x x
    +
    819  // 7 5 3 x x
    +
    820  // 8 6 4 2 x
    +
    821  // 9 7 5 3 1
    +
    822  //
    +
    823  // Elements of the same epoch are somehow "independent" and so they can potentially run in parallel,
    +
    824  // given that previous epoch has been completed. Since scheduling happens sequentially, elements
    +
    825  // of the same epoch will be ordered starting from top-most one, resulting in
    +
    826  //
    +
    827  // 7 x x x x
    +
    828  // 10 5 x x x
    +
    829  // 12 8 3 x x
    +
    830  // 14 11 6 2 x
    +
    831  // 15 13 9 4 1
    +
    832  for (SizeType k = idx_last_sweep_b; k > -maxsteps_b; --k) {
    +
    833  auto& mat_t = t_panels.nextResource();
    +
    834  auto& panel_hh = hh_panels.nextResource();
    +
    835  auto& mat_v = v_panels.nextResource();
    +
    836  auto& mat_w = w_panels.nextResource();
    +
    837  auto& mat_w2 = w2_panels.nextResource();
    +
    838  auto& mat_w2tmp = w2tmp_panels.nextResource();
    +
    839 
    +
    840  for (SizeType i_b = std::abs<SizeType>(k), j_b = std::max<SizeType>(0, k);
    +
    841  i_b < j_b + nrStepsForSweep(j_b * b, mat_hh.size().cols(), b); i_b += 2, ++j_b) {
    +
    842  const SizeType step_b = i_b - j_b;
    +
    843  const GlobalElementIndex ij_el(i_b * b, j_b * b);
    +
    844  const GlobalTileIndex ij_g(dist_hh.globalTileIndex(ij_el));
    +
    845 
    +
    846  const comm::Index2D rank = dist_hh.rankIndex();
    +
    847  const comm::Index2D rankHH = dist_hh.rankGlobalTile(ij_g);
    +
    848 
    +
    849  // Note: reflector with size = 1 must be ignored, except for the last step of the last sweep
    +
    850  // with complex type
    +
    851  const SizeType nrefls = [&]() {
    +
    852  const bool allowSize1 = isComplex_v<T> && j_b == idx_last_sweep_b &&
    +
    853  step_b == nrStepsForSweep(j_b * b, mat_hh.size().cols(), b) - 1;
    +
    854  const GlobalElementSize delta(dist_hh.size().rows() - ij_el.row() - 1,
    +
    855  std::min(b, dist_hh.size().cols() - ij_el.col()));
    +
    856  return std::min(b, std::min(delta.rows() - (allowSize1 ? 0 : 1), delta.cols()));
    +
    857  }();
    +
    858 
    +
    859  const TileAccessHelper helper(b, nrefls, dist_hh, dist_e_rt, ij_el);
    +
    860  const DistIndexing indexing_helper(helper, dist_hh, b, ij_g, i_b);
    +
    861 
    +
    862  if (!indexing_helper.isInvolved())
    +
    863  continue;
    +
    864 
    +
    865  if (nrefls < b) {
    +
    866  mat_t.setWidth(nrefls);
    +
    867  mat_v.setWidth(nrefls);
    +
    868  mat_w.setWidth(nrefls);
    +
    869  mat_w2.setHeight(nrefls);
    +
    870  mat_w2tmp.setHeight(nrefls);
    +
    871  }
    +
    872 
    +
    873  // Note:
    +
    874  // From HH it is possible to extract V that is needed for computing W and W2, both required
    +
    875  // for updating E.
    +
    876 
    +
    877  // Send HH to all involved ranks: broadcast on row + send p2p on col
    +
    878  const LocalTileIndex ij_hh_panel = indexing_helper.wsIndexHH();
    +
    879 
    +
    880  // Broadcast on ROW
    +
    881  if (grid.size().cols() > 1 && rank.row() == rankHH.row()) {
    +
    882  if (rank.col() == rankHH.col()) {
    +
    883  ex::start_detached(comm::scheduleSendBcast(
    +
    884  mpi_chain_row(), splitTile(mat_hh.read(ij_g), helper.specHHCompact())));
    +
    885  }
    +
    886  else {
    +
    887  ex::start_detached(comm::scheduleRecvBcast(mpi_chain_row(), rankHH.col(),
    +
    888  splitTile(panel_hh.readwrite(ij_hh_panel),
    +
    889  helper.specHHCompact(true))));
    +
    890  }
    +
    891  }
    +
    892 
    +
    893  // Jump to the next loop iteration if the local part of E is not affected by the update.
    +
    894  const SizeType ncols_local = dist_e_rt.localNrTiles().cols();
    +
    895  if (ncols_local == 0)
    +
    896  continue;
    +
    897 
    +
    898  // Send P2P on col
    +
    899  if (helper.affectsMultipleRanks()) {
    +
    900  const comm::IndexT_MPI rank_src = rankHH.row();
    +
    901  const comm::IndexT_MPI rank_dst = indexing_helper.rankRowPartner();
    +
    902 
    +
    903  if (rank.row() == rank_src) {
    +
    904  auto tile_hh = rank.col() == rankHH.col()
    +
    905  ? splitTile(mat_hh.read(ij_g), helper.specHHCompact())
    +
    906  : splitTile(panel_hh.read(ij_hh_panel), helper.specHHCompact(true));
    +
    907  ex::start_detached(comm::scheduleSend(mpi_chain_col(), rank_dst, 0, std::move(tile_hh)));
    +
    908  }
    +
    909  else if (rank.row() == rank_dst) {
    +
    910  ex::start_detached(comm::scheduleRecv(mpi_chain_col(), rank_src, 0,
    +
    911  panel_hh.readwrite(ij_hh_panel)));
    +
    912  }
    +
    913  }
    +
    914 
    +
    915  // Batched application increases the communication messages and volume,
    +
    916  // therefore we use it only for local applications.
    +
    917  const SizeType current_group_size = helper.affectsMultipleRanks() ? b : group_size;
    +
    918 
    +
    919  // COMPUTE V and W from HH and T
    +
    920  auto tile_hh = (rankHH == rank)
    +
    921  ? splitTile(mat_hh.read(ij_g), helper.specHHCompact())
    +
    922  : splitTile(panel_hh.read(ij_hh_panel), helper.specHHCompact(true));
    +
    923  auto [tile_v_unshared, tile_w_unshared] =
    +
    924  helperBackend.computeVW(current_group_size, indexing_helper.wsIndexHH(), helper,
    +
    925  std::move(tile_hh), mat_v, mat_t, mat_w);
    +
    926  auto tile_v = matrix::shareReadWriteTile(ex::make_unique_any_sender(std::move(tile_v_unshared)));
    +
    927  auto tile_w = matrix::shareReadWriteTile(ex::make_unique_any_sender(std::move(tile_w_unshared)));
    +
    928 
    +
    929  // UPDATE E
    +
    930  for (SizeType j_e = 0; j_e < ncols_local; ++j_e) {
    +
    931  const SizeType j_e_g = dist_e_rt.template globalTileFromLocalTile<Coord::Col>(j_e);
    +
    932  const LocalTileIndex idx_w2(indexing_helper.wsIndexHH().row(), j_e);
    +
    933 
    +
    934  const GlobalTileIndex idx_e_top = helper.topIndexE(j_e_g);
    +
    935  const auto nb = mat_e_rt.tileSize(idx_e_top).cols();
    +
    936 
    +
    937  // SINGLE ROW UPDATE
    +
    938  if (!helper.affectsMultipleTiles()) {
    +
    939  ex::start_detached(ex::when_all(ex::just(current_group_size), tile_v, tile_w,
    +
    940  splitTile(mat_w2.readwrite(idx_w2), helper.specW2(nb)),
    +
    941  mat_e_rt.readwrite(idx_e_top)) |
    +
    942  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    +
    943  dlaf::internal::Policy<B>(thread_priority::normal),
    +
    944  ApplyHHToSingleTileRow<B, T>{}));
    +
    945  }
    +
    946  // TWO ROWs
    +
    947  else {
    +
    948  const GlobalTileIndex idx_e_bottom = helper.bottomIndexE(j_e_g);
    +
    949 
    +
    950  // TWO ROWs (same RANK)
    +
    951  if (!helper.affectsMultipleRanks()) {
    +
    952  ex::start_detached(
    +
    953  ex::when_all(ex::just(current_group_size), tile_v, tile_w,
    +
    954  splitTile(mat_w2.readwrite(idx_w2), helper.specW2(nb)),
    +
    955  mat_e_rt.readwrite(idx_e_top), mat_e_rt.readwrite(idx_e_bottom)) |
    +
    956  dlaf::internal::transform<dlaf::internal::TransformDispatchType::Blas>(
    +
    957  dlaf::internal::Policy<B>(thread_priority::normal), ApplyHHToDoubleTileRow<B, T>{}));
    +
    958  }
    +
    959  // TWO ROWs TWO RANKs UPDATE (MAIN + PARTNER)
    +
    960  else {
    +
    961  const bool is_top_rank = rank.row() == rankHH.row();
    +
    962  const comm::IndexT_MPI rank_partner =
    +
    963  is_top_rank ? indexing_helper.rankRowPartner() : rankHH.row();
    +
    964 
    +
    965  const comm::IndexT_MPI tag = to_int(j_e + i_b * ncols_local);
    +
    966 
    +
    967  const matrix::SubTileSpec spec_vw = helper.specHH(is_top_rank);
    +
    968 
    +
    969  const auto idx_e = is_top_rank ? idx_e_top : idx_e_bottom;
    +
    970  const auto sz_e = mat_e_rt.tileSize(idx_e);
    +
    971  const matrix::SubTileSpec spec_e{{(is_top_rank ? 1 : 0), 0},
    +
    972  {sz_e.rows() - (is_top_rank ? 1 : 0), sz_e.cols()}};
    +
    973 
    +
    974  auto subtile_v = splitTile(tile_v, spec_vw);
    +
    975  auto subtile_w = splitTile(tile_w, spec_vw);
    +
    976  auto subtile_e_ro = splitTile(mat_e_rt.read(idx_e), spec_e);
    +
    977 
    +
    978  // W2 = V* . E
    +
    979  ex::start_detached(
    +
    980  dlaf::internal::whenAllLift(blas::Op::ConjTrans, blas::Op::NoTrans, T(1),
    +
    981  std::move(subtile_v), std::move(subtile_e_ro), T(0),
    +
    982  splitTile(mat_w2tmp.readwrite(idx_w2), helper.specW2(nb))) |
    +
    983  dlaf::tile::gemm(dlaf::internal::Policy<B>(thread_priority::normal)));
    +
    984 
    +
    985  // Compute final W2 by adding the contribution from the partner rank
    +
    986  ex::start_detached( //
    +
    987  comm::scheduleAllSumP2P<B>(mpi_col_comm, rank_partner, tag,
    +
    988  splitTile(mat_w2tmp.read(idx_w2), helper.specW2(nb)),
    +
    989  splitTile(mat_w2.readwrite(idx_w2), helper.specW2(nb))));
    +
    990 
    +
    991  auto subtile_e = splitTile(mat_e_rt.readwrite(idx_e), spec_e);
    +
    992  // E -= W . W2
    +
    993  ex::start_detached( //
    +
    994  dlaf::internal::whenAllLift(blas::Op::NoTrans, blas::Op::NoTrans, T(-1),
    +
    995  std::move(subtile_w),
    +
    996  splitTile(mat_w2.read(idx_w2), helper.specW2(nb)), T(1),
    +
    997  std::move(subtile_e)) |
    +
    998  dlaf::tile::gemm(dlaf::internal::Policy<B>(thread_priority::normal)));
    +
    999  }
    +
    1000  }
    +
    1001  }
    +
    1002 
    +
    1003  mat_t.reset();
    +
    1004  panel_hh.reset();
    +
    1005  mat_v.reset();
    +
    1006  mat_w.reset();
    +
    1007  mat_w2tmp.reset();
    +
    1008  mat_w2.reset();
    +
    1009  }
    +
    1010  }
    +
    1011 }
    1012 }
    -
    1013 }
    @@ -1083,7 +1082,6 @@
    dlaf::matrix::Distribution::nrTiles
    const GlobalTileSize & nrTiles() const noexcept
    Returns the number of tiles of the global matrix (2D size).
    Definition: distribution.h:142
    dlaf::matrix::Distribution::rankGlobalTile
    comm::Index2D rankGlobalTile(const GlobalTileIndex &global_tile) const noexcept
    Definition: distribution.h:210
    dlaf::matrix::Distribution::globalTileFromGlobalElement
    SizeType globalTileFromGlobalElement(SizeType global_element) const noexcept
    Definition: distribution.h:292
    -
    dlaf::matrix::RetiledMatrix
    Definition: retiled_matrix.h:36
    dlaf::matrix::Tile
    Definition: tile.h:329
    dlaf::comm::IndexT_MPI
    int IndexT_MPI
    Type used for indexes in MPI API.
    Definition: communicator.h:23
    @@ -1101,16 +1099,15 @@ -
    retiled_matrix.h
    round_robin.h
    single_threaded_blas.h
    dlaf::common::RoundRobin
    Definition: round_robin.h:20
    -
    dlaf::eigensolver::internal::bt_tridiag::ApplyHHToDoubleTileRow
    Definition: impl.h:259
    -
    dlaf::eigensolver::internal::bt_tridiag::ApplyHHToSingleTileRow
    Definition: impl.h:174
    -
    dlaf::eigensolver::internal::bt_tridiag::DistIndexing
    Definition: impl.h:446
    -
    dlaf::eigensolver::internal::bt_tridiag::HHManager
    Definition: impl.h:501
    -
    dlaf::eigensolver::internal::bt_tridiag::TileAccessHelper
    Definition: impl.h:322
    +
    dlaf::eigensolver::internal::bt_tridiag::ApplyHHToDoubleTileRow
    Definition: impl.h:258
    +
    dlaf::eigensolver::internal::bt_tridiag::ApplyHHToSingleTileRow
    Definition: impl.h:173
    +
    dlaf::eigensolver::internal::bt_tridiag::DistIndexing
    Definition: impl.h:445
    +
    dlaf::eigensolver::internal::bt_tridiag::HHManager
    Definition: impl.h:500
    +
    dlaf::eigensolver::internal::bt_tridiag::TileAccessHelper
    Definition: impl.h:321
    dlaf::matrix::Panel
    Definition: panel.h:552
    dlaf::matrix::Panel::readwrite
    ReadWriteSenderType readwrite(LocalTileIndex index)
    Definition: panel.h:570
    dlaf::matrix::SubTileSpec
    Contains the information to create a subtile.
    Definition: tile.h:109
    diff --git a/master/eigensolver_2eigensolver_8h_source.html b/master/eigensolver_2eigensolver_8h_source.html index d2d4c70104..e97112b26d 100644 --- a/master/eigensolver_2eigensolver_8h_source.html +++ b/master/eigensolver_2eigensolver_8h_source.html @@ -106,54 +106,60 @@
    50  DLAF_ASSERT(square_blocksize(eigenvectors), eigenvectors);
    51  DLAF_ASSERT(eigenvectors.size() == mat.size(), eigenvectors, mat);
    52  DLAF_ASSERT(eigenvectors.blockSize() == mat.blockSize(), eigenvectors, mat);
    -
    53 
    -
    54  internal::Eigensolver<B, D, T>::call(uplo, mat, eigenvalues, eigenvectors);
    -
    55 }
    +
    53  DLAF_ASSERT(single_tile_per_block(mat), mat);
    +
    54  DLAF_ASSERT(single_tile_per_block(eigenvalues), eigenvalues);
    +
    55  DLAF_ASSERT(single_tile_per_block(eigenvectors), eigenvectors);
    56 
    -
    69 template <Backend B, Device D, class T>
    -
    70 EigensolverResult<T, D> eigensolver(blas::Uplo uplo, Matrix<T, D>& mat) {
    -
    71  const SizeType size = mat.size().rows();
    -
    72  matrix::Matrix<BaseType<T>, D> eigenvalues(LocalElementSize(size, 1),
    -
    73  TileElementSize(mat.blockSize().rows(), 1));
    -
    74  matrix::Matrix<T, D> eigenvectors(LocalElementSize(size, size), mat.blockSize());
    -
    75 
    -
    76  eigensolver<B, D, T>(uplo, mat, eigenvalues, eigenvectors);
    -
    77  return {std::move(eigenvalues), std::move(eigenvectors)};
    -
    78 }
    -
    79 
    -
    95 template <Backend B, Device D, class T>
    -
    96 void eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, D>& mat,
    -
    97  Matrix<BaseType<T>, D>& eigenvalues, Matrix<T, D>& eigenvectors) {
    -
    98  DLAF_ASSERT(matrix::equal_process_grid(mat, grid), mat);
    -
    99  DLAF_ASSERT(square_size(mat), mat);
    -
    100  DLAF_ASSERT(square_blocksize(mat), mat);
    -
    101  DLAF_ASSERT(matrix::local_matrix(eigenvalues), eigenvalues);
    -
    102  DLAF_ASSERT(eigenvalues.size().rows() == eigenvectors.size().rows(), eigenvalues, eigenvectors);
    -
    103  DLAF_ASSERT(eigenvalues.blockSize().rows() == eigenvectors.blockSize().rows(), eigenvalues,
    -
    104  eigenvectors);
    -
    105  DLAF_ASSERT(matrix::equal_process_grid(eigenvectors, grid), eigenvectors);
    -
    106  DLAF_ASSERT(square_size(eigenvectors), eigenvectors);
    -
    107  DLAF_ASSERT(square_blocksize(eigenvectors), eigenvectors);
    -
    108  DLAF_ASSERT(eigenvectors.size() == mat.size(), eigenvectors, mat);
    -
    109  DLAF_ASSERT(eigenvectors.blockSize() == mat.blockSize(), eigenvectors, mat);
    -
    110 
    -
    111  internal::Eigensolver<B, D, T>::call(grid, uplo, mat, eigenvalues, eigenvectors);
    -
    112 }
    -
    113 
    -
    127 template <Backend B, Device D, class T>
    -
    128 EigensolverResult<T, D> eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, D>& mat) {
    -
    129  const SizeType size = mat.size().rows();
    -
    130  matrix::Matrix<BaseType<T>, D> eigenvalues(LocalElementSize(size, 1),
    -
    131  TileElementSize(mat.blockSize().rows(), 1));
    -
    132  matrix::Matrix<T, D> eigenvectors(GlobalElementSize(size, size), mat.blockSize(), grid);
    -
    133 
    -
    134  eigensolver<B, D, T>(grid, uplo, mat, eigenvalues, eigenvectors);
    -
    135  return {std::move(eigenvalues), std::move(eigenvectors)};
    -
    136 }
    -
    137 }
    +
    57  internal::Eigensolver<B, D, T>::call(uplo, mat, eigenvalues, eigenvectors);
    +
    58 }
    +
    59 
    +
    72 template <Backend B, Device D, class T>
    +
    73 EigensolverResult<T, D> eigensolver(blas::Uplo uplo, Matrix<T, D>& mat) {
    +
    74  const SizeType size = mat.size().rows();
    +
    75  matrix::Matrix<BaseType<T>, D> eigenvalues(LocalElementSize(size, 1),
    +
    76  TileElementSize(mat.blockSize().rows(), 1));
    +
    77  matrix::Matrix<T, D> eigenvectors(LocalElementSize(size, size), mat.blockSize());
    +
    78 
    +
    79  eigensolver<B, D, T>(uplo, mat, eigenvalues, eigenvectors);
    +
    80  return {std::move(eigenvalues), std::move(eigenvectors)};
    +
    81 }
    +
    82 
    +
    98 template <Backend B, Device D, class T>
    +
    99 void eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, D>& mat,
    +
    100  Matrix<BaseType<T>, D>& eigenvalues, Matrix<T, D>& eigenvectors) {
    +
    101  DLAF_ASSERT(matrix::equal_process_grid(mat, grid), mat);
    +
    102  DLAF_ASSERT(square_size(mat), mat);
    +
    103  DLAF_ASSERT(square_blocksize(mat), mat);
    +
    104  DLAF_ASSERT(matrix::local_matrix(eigenvalues), eigenvalues);
    +
    105  DLAF_ASSERT(eigenvalues.size().rows() == eigenvectors.size().rows(), eigenvalues, eigenvectors);
    +
    106  DLAF_ASSERT(eigenvalues.blockSize().rows() == eigenvectors.blockSize().rows(), eigenvalues,
    +
    107  eigenvectors);
    +
    108  DLAF_ASSERT(matrix::equal_process_grid(eigenvectors, grid), eigenvectors);
    +
    109  DLAF_ASSERT(square_size(eigenvectors), eigenvectors);
    +
    110  DLAF_ASSERT(square_blocksize(eigenvectors), eigenvectors);
    +
    111  DLAF_ASSERT(eigenvectors.size() == mat.size(), eigenvectors, mat);
    +
    112  DLAF_ASSERT(eigenvectors.blockSize() == mat.blockSize(), eigenvectors, mat);
    +
    113  DLAF_ASSERT(single_tile_per_block(mat), mat);
    +
    114  DLAF_ASSERT(single_tile_per_block(eigenvalues), eigenvalues);
    +
    115  DLAF_ASSERT(single_tile_per_block(eigenvectors), eigenvectors);
    +
    116 
    +
    117  internal::Eigensolver<B, D, T>::call(grid, uplo, mat, eigenvalues, eigenvectors);
    +
    118 }
    +
    119 
    +
    133 template <Backend B, Device D, class T>
    +
    134 EigensolverResult<T, D> eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, D>& mat) {
    +
    135  const SizeType size = mat.size().rows();
    +
    136  matrix::Matrix<BaseType<T>, D> eigenvalues(LocalElementSize(size, 1),
    +
    137  TileElementSize(mat.blockSize().rows(), 1));
    +
    138  matrix::Matrix<T, D> eigenvectors(GlobalElementSize(size, size), mat.blockSize(), grid);
    +
    139 
    +
    140  eigensolver<B, D, T>(grid, uplo, mat, eigenvalues, eigenvectors);
    +
    141  return {std::move(eigenvalues), std::move(eigenvectors)};
    +
    142 }
    +
    143 }
    dlaf::comm::CommunicatorGrid
    Definition: communicator_grid.h:42
    -
    dlaf::matrix::Matrix
    Definition: matrix.h:55
    +
    dlaf::matrix::Matrix
    Definition: matrix.h:56
    communicator_grid.h
    dlaf::eigensolver::eigensolver
    void eigensolver(blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)
    Definition: eigensolver.h:39
    diff --git a/master/eigensolver_2reduction__to__band_2impl_8h_source.html b/master/eigensolver_2reduction__to__band_2impl_8h_source.html index fc82107637..e0d7136b66 100644 --- a/master/eigensolver_2reduction__to__band_2impl_8h_source.html +++ b/master/eigensolver_2reduction__to__band_2impl_8h_source.html @@ -113,1361 +113,1359 @@
    42 #include <dlaf/matrix/index.h>
    43 #include <dlaf/matrix/matrix.h>
    44 #include <dlaf/matrix/panel.h>
    -
    45 #include <dlaf/matrix/retiled_matrix.h>
    -
    46 #include <dlaf/matrix/tile.h>
    -
    47 #include <dlaf/matrix/views.h>
    -
    48 #include <dlaf/schedulers.h>
    -
    49 #include <dlaf/sender/traits.h>
    -
    50 #include <dlaf/types.h>
    -
    51 #include <dlaf/util_math.h>
    -
    52 #include <dlaf/util_matrix.h>
    -
    53 
    -
    54 namespace dlaf::eigensolver::internal {
    -
    55 
    -
    56 // Given a vector of vectors, reduce all vectors in the first one using sum operation
    -
    57 template <class T>
    -
    58 void reduceColumnVectors(std::vector<common::internal::vector<T>>& columnVectors) {
    -
    59  for (std::size_t i = 1; i < columnVectors.size(); ++i) {
    -
    60  DLAF_ASSERT_HEAVY(columnVectors[0].size() == columnVectors[i].size(), columnVectors[0].size(),
    -
    61  columnVectors[i].size());
    -
    62  for (SizeType j = 0; j < columnVectors[0].size(); ++j)
    -
    63  columnVectors[0][j] += columnVectors[i][j];
    -
    64  }
    -
    65 }
    -
    66 
    -
    67 namespace red2band {
    -
    68 
    -
    69 // Extract x0 and compute local cumulative sum of squares of the reflector column
    -
    70 template <Device D, class T>
    -
    71 std::array<T, 2> computeX0AndSquares(const bool has_head, const std::vector<matrix::Tile<T, D>>& panel,
    -
    72  SizeType j) {
    -
    73  std::array<T, 2> x0_and_squares{0, 0};
    -
    74  auto it_begin = panel.begin();
    -
    75  auto it_end = panel.end();
    -
    76 
    -
    77  common::internal::SingleThreadedBlasScope single;
    -
    78 
    -
    79  if (has_head) {
    -
    80  auto& tile_v0 = *it_begin++;
    -
    81 
    -
    82  const TileElementIndex idx_x0(j, j);
    -
    83  x0_and_squares[0] = tile_v0(idx_x0);
    -
    84 
    -
    85  T* reflector_ptr = tile_v0.ptr({idx_x0});
    -
    86  x0_and_squares[1] =
    -
    87  blas::dot(tile_v0.size().rows() - idx_x0.row(), reflector_ptr, 1, reflector_ptr, 1);
    -
    88  }
    -
    89 
    -
    90  for (auto it = it_begin; it != it_end; ++it) {
    -
    91  const auto& tile = *it;
    -
    92 
    -
    93  T* reflector_ptr = tile.ptr({0, j});
    -
    94  x0_and_squares[1] += blas::dot(tile.size().rows(), reflector_ptr, 1, reflector_ptr, 1);
    -
    95  }
    -
    96  return x0_and_squares;
    -
    97 }
    -
    98 
    -
    99 template <Device D, class T>
    -
    100 T computeReflectorAndTau(const bool has_head, const std::vector<matrix::Tile<T, D>>& panel,
    -
    101  const SizeType j, std::array<T, 2> x0_and_squares) {
    -
    102  const T norm = std::sqrt(x0_and_squares[1]);
    -
    103  const T x0 = x0_and_squares[0];
    -
    104  const T y = std::signbit(std::real(x0_and_squares[0])) ? norm : -norm;
    -
    105  const T tau = (y - x0) / y;
    -
    106 
    -
    107  auto it_begin = panel.begin();
    -
    108  auto it_end = panel.end();
    -
    109 
    -
    110  common::internal::SingleThreadedBlasScope single;
    -
    111 
    -
    112  if (has_head) {
    -
    113  const auto& tile_v0 = *it_begin++;
    -
    114 
    -
    115  const TileElementIndex idx_x0(j, j);
    -
    116  tile_v0(idx_x0) = y;
    -
    117 
    -
    118  if (j + 1 < tile_v0.size().rows()) {
    -
    119  T* v = tile_v0.ptr({j + 1, j});
    -
    120  blas::scal(tile_v0.size().rows() - (j + 1), T(1) / (x0 - y), v, 1);
    -
    121  }
    -
    122  }
    -
    123 
    -
    124  for (auto it = it_begin; it != it_end; ++it) {
    -
    125  auto& tile_v = *it;
    -
    126  T* v = tile_v.ptr({0, j});
    -
    127  blas::scal(tile_v.size().rows(), T(1) / (x0 - y), v, 1);
    -
    128  }
    -
    129 
    -
    130  return tau;
    -
    131 }
    -
    132 
    -
    133 template <Device D, class T>
    -
    134 void computeWTrailingPanel(const bool has_head, const std::vector<matrix::Tile<T, D>>& panel,
    -
    135  common::internal::vector<T>& w, SizeType j, const SizeType pt_cols,
    -
    136  const std::size_t begin, const std::size_t end) {
    -
    137  // for each tile in the panel, consider just the trailing panel
    -
    138  // i.e. all rows (height = reflector), just columns to the right of the current reflector
    -
    139  if (!(pt_cols > 0))
    -
    140  return;
    -
    141 
    -
    142  const TileElementIndex index_el_x0(j, j);
    -
    143  bool has_first_component = has_head;
    -
    144 
    -
    145  common::internal::SingleThreadedBlasScope single;
    -
    146 
    -
    147  // W = Pt* . V
    -
    148  for (auto index = begin; index < end; ++index) {
    -
    149  const matrix::Tile<const T, D>& tile_a = panel[index];
    -
    150  const SizeType first_element = has_first_component ? index_el_x0.row() : 0;
    -
    151 
    -
    152  TileElementIndex pt_start{first_element, index_el_x0.col() + 1};
    -
    153  TileElementSize pt_size{tile_a.size().rows() - pt_start.row(), pt_cols};
    -
    154  TileElementIndex v_start{first_element, index_el_x0.col()};
    -
    155 
    -
    156  if (has_first_component) {
    -
    157  const TileElementSize offset{1, 0};
    -
    158 
    -
    159  const T fake_v = 1;
    -
    160  blas::gemv(blas::Layout::ColMajor, blas::Op::ConjTrans, offset.rows(), pt_size.cols(), T(1),
    -
    161  tile_a.ptr(pt_start), tile_a.ld(), &fake_v, 1, T(0), w.data(), 1);
    -
    162 
    -
    163  pt_start = pt_start + offset;
    -
    164  v_start = v_start + offset;
    -
    165  pt_size = pt_size - offset;
    -
    166 
    -
    167  has_first_component = false;
    -
    168  }
    -
    169 
    -
    170  if (pt_start.isIn(tile_a.size())) {
    -
    171  // W += 1 . A* . V
    -
    172  blas::gemv(blas::Layout::ColMajor, blas::Op::ConjTrans, pt_size.rows(), pt_size.cols(), T(1),
    -
    173  tile_a.ptr(pt_start), tile_a.ld(), tile_a.ptr(v_start), 1, T(1), w.data(), 1);
    -
    174  }
    -
    175  }
    -
    176 }
    -
    177 
    -
    178 template <Device D, class T>
    -
    179 void updateTrailingPanel(const bool has_head, const std::vector<matrix::Tile<T, D>>& panel, SizeType j,
    -
    180  const std::vector<T>& w, const T tau, const std::size_t begin,
    -
    181  const std::size_t end) {
    -
    182  const TileElementIndex index_el_x0(j, j);
    -
    183 
    -
    184  bool has_first_component = has_head;
    -
    185 
    -
    186  common::internal::SingleThreadedBlasScope single;
    -
    187 
    -
    188  // GER Pt = Pt - tau . v . w*
    -
    189  for (auto index = begin; index < end; ++index) {
    -
    190  const matrix::Tile<T, D>& tile_a = panel[index];
    -
    191  const SizeType first_element = has_first_component ? index_el_x0.row() : 0;
    -
    192 
    -
    193  TileElementIndex pt_start{first_element, index_el_x0.col() + 1};
    -
    194  TileElementSize pt_size{tile_a.size().rows() - pt_start.row(),
    -
    195  tile_a.size().cols() - pt_start.col()};
    -
    196  TileElementIndex v_start{first_element, index_el_x0.col()};
    -
    197 
    -
    198  if (has_first_component) {
    -
    199  const TileElementSize offset{1, 0};
    -
    200 
    -
    201  // Pt = Pt - tau * v[0] * w*
    -
    202  const T fake_v = 1;
    -
    203  blas::ger(blas::Layout::ColMajor, 1, pt_size.cols(), -dlaf::conj(tau), &fake_v, 1, w.data(), 1,
    -
    204  tile_a.ptr(pt_start), tile_a.ld());
    -
    205 
    -
    206  pt_start = pt_start + offset;
    -
    207  v_start = v_start + offset;
    -
    208  pt_size = pt_size - offset;
    -
    209 
    -
    210  has_first_component = false;
    -
    211  }
    -
    212 
    -
    213  if (pt_start.isIn(tile_a.size())) {
    -
    214  // Pt = Pt - tau * v * w*
    -
    215  blas::ger(blas::Layout::ColMajor, pt_size.rows(), pt_size.cols(), -dlaf::conj(tau),
    -
    216  tile_a.ptr(v_start), 1, w.data(), 1, tile_a.ptr(pt_start), tile_a.ld());
    -
    217  }
    -
    218  }
    -
    219 }
    -
    220 
    -
    221 template <Backend B, typename ASender, typename WSender, typename XSender>
    -
    222 void hemmDiag(pika::execution::thread_priority priority, ASender&& tile_a, WSender&& tile_w,
    -
    223  XSender&& tile_x) {
    -
    224  using T = dlaf::internal::SenderElementType<ASender>;
    -
    225  pika::execution::experimental::start_detached(
    -
    226  dlaf::internal::whenAllLift(blas::Side::Left, blas::Uplo::Lower, T(1),
    -
    227  std::forward<ASender>(tile_a), std::forward<WSender>(tile_w), T(1),
    -
    228  std::forward<XSender>(tile_x)) |
    -
    229  tile::hemm(dlaf::internal::Policy<B>(priority)));
    -
    230 }
    -
    231 
    -
    232 // X += op(A) * W
    -
    233 template <Backend B, typename ASender, typename WSender, typename XSender>
    -
    234 void hemmOffDiag(pika::execution::thread_priority priority, blas::Op op, ASender&& tile_a,
    -
    235  WSender&& tile_w, XSender&& tile_x) {
    -
    236  using T = dlaf::internal::SenderElementType<ASender>;
    -
    237  pika::execution::experimental::start_detached(
    -
    238  dlaf::internal::whenAllLift(op, blas::Op::NoTrans, T(1), std::forward<ASender>(tile_a),
    -
    239  std::forward<WSender>(tile_w), T(1), std::forward<XSender>(tile_x)) |
    -
    240  tile::gemm(dlaf::internal::Policy<B>(priority)));
    -
    241 }
    -
    242 
    -
    243 template <Backend B, typename VSender, typename XSender, typename ASender>
    -
    244 void her2kDiag(pika::execution::thread_priority priority, VSender&& tile_v, XSender&& tile_x,
    -
    245  ASender&& tile_a) {
    -
    246  using T = dlaf::internal::SenderElementType<VSender>;
    -
    247  pika::execution::experimental::start_detached(
    -
    248  dlaf::internal::whenAllLift(blas::Uplo::Lower, blas::Op::NoTrans, T(-1),
    -
    249  std::forward<VSender>(tile_v), std::forward<XSender>(tile_x),
    -
    250  BaseType<T>(1), std::forward<ASender>(tile_a)) |
    -
    251  tile::her2k(dlaf::internal::Policy<B>(priority)));
    -
    252 }
    -
    253 
    -
    254 // C -= A . B*
    -
    255 template <Backend B, typename ASender, typename BSender, typename CSender>
    -
    256 void her2kOffDiag(pika::execution::thread_priority priority, ASender&& tile_a, BSender&& tile_b,
    -
    257  CSender&& tile_c) {
    -
    258  using T = dlaf::internal::SenderElementType<ASender>;
    -
    259  pika::execution::experimental::start_detached(
    -
    260  dlaf::internal::whenAllLift(blas::Op::NoTrans, blas::Op::ConjTrans, T(-1),
    -
    261  std::forward<ASender>(tile_a), std::forward<BSender>(tile_b), T(1),
    -
    262  std::forward<CSender>(tile_c)) |
    -
    263  tile::gemm(dlaf::internal::Policy<B>(priority)));
    -
    264 }
    -
    265 
    -
    266 namespace local {
    -
    267 
    -
    268 template <Device D, class T>
    -
    269 T computeReflector(const std::vector<matrix::Tile<T, D>>& panel, SizeType j) {
    -
    270  constexpr bool has_head = true;
    -
    271 
    -
    272  std::array<T, 2> x0_and_squares = computeX0AndSquares(has_head, panel, j);
    -
    273 
    -
    274  auto tau = computeReflectorAndTau(has_head, panel, j, std::move(x0_and_squares));
    -
    275 
    -
    276  return tau;
    -
    277 }
    -
    278 
    -
    279 template <class MatrixLikeA, class MatrixLikeTaus>
    -
    280 void computePanelReflectors(MatrixLikeA& mat_a, MatrixLikeTaus& mat_taus, const SizeType j_sub,
    -
    281  const matrix::SubPanelView& panel_view) {
    -
    282  static Device constexpr D = MatrixLikeA::device;
    -
    283  using T = typename MatrixLikeA::ElementType;
    -
    284  namespace ex = pika::execution::experimental;
    -
    285  namespace di = dlaf::internal;
    -
    286 
    -
    287  std::vector<matrix::ReadWriteTileSender<T, D>> panel_tiles;
    -
    288  panel_tiles.reserve(to_sizet(std::distance(panel_view.iteratorLocal().begin(),
    -
    289  panel_view.iteratorLocal().end())));
    -
    290  for (const auto& i : panel_view.iteratorLocal()) {
    -
    291  const matrix::SubTileSpec& spec = panel_view(i);
    -
    292  panel_tiles.emplace_back(matrix::splitTile(mat_a.readwrite(i), spec));
    -
    293  }
    -
    294 
    -
    295  const std::size_t nthreads = getReductionToBandPanelNWorkers();
    -
    296  auto s =
    -
    297  ex::when_all(ex::just(std::make_unique<pika::barrier<>>(nthreads),
    -
    298  std::vector<common::internal::vector<T>>{}), // w (internally required)
    -
    299  mat_taus.readwrite(LocalTileIndex(j_sub, 0)),
    -
    300  ex::when_all_vector(std::move(panel_tiles))) |
    -
    301  ex::transfer(di::getBackendScheduler<Backend::MC>(pika::execution::thread_priority::high)) |
    -
    302  ex::bulk(nthreads, [nthreads, cols = panel_view.cols()](const std::size_t index, auto& barrier_ptr,
    -
    303  auto& w, auto& taus, auto& tiles) {
    -
    304  const auto barrier_busy_wait = getReductionToBandBarrierBusyWait();
    -
    305  const std::size_t batch_size = util::ceilDiv(tiles.size(), nthreads);
    -
    306  const std::size_t begin = index * batch_size;
    -
    307  const std::size_t end = std::min(index * batch_size + batch_size, tiles.size());
    -
    308  const SizeType nrefls = taus.size().rows();
    -
    309 
    -
    310  if (index == 0) {
    -
    311  w.resize(nthreads);
    -
    312  }
    -
    313 
    -
    314  for (SizeType j = 0; j < nrefls; ++j) {
    -
    315  // STEP1: compute tau and reflector (single-thread)
    -
    316  if (index == 0) {
    -
    317  taus({j, 0}) = computeReflector(tiles, j);
    -
    318  }
    -
    319 
    -
    320  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    -
    321 
    -
    322  // STEP2a: compute w (multi-threaded)
    -
    323  const SizeType pt_cols = cols - (j + 1);
    -
    324  if (pt_cols == 0)
    -
    325  break;
    -
    326  const bool has_head = (index == 0);
    -
    327 
    -
    328  w[index] = common::internal::vector<T>(pt_cols, 0);
    -
    329  computeWTrailingPanel(has_head, tiles, w[index], j, pt_cols, begin, end);
    -
    330  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    -
    331 
    -
    332  // STEP2b: reduce w results (single-threaded)
    -
    333  if (index == 0)
    -
    334  dlaf::eigensolver::internal::reduceColumnVectors(w);
    -
    335  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    -
    336 
    -
    337  // STEP3: update trailing panel (multi-threaded)
    -
    338  updateTrailingPanel(has_head, tiles, j, w[0], taus({j, 0}), begin, end);
    -
    339  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    -
    340  }
    -
    341  });
    -
    342  ex::start_detached(std::move(s));
    -
    343 }
    -
    344 
    -
    345 template <Backend B, Device D, class T>
    -
    346 void setupReflectorPanelV(bool has_head, const matrix::SubPanelView& panel_view, const SizeType nrefls,
    -
    347  matrix::Panel<Coord::Col, T, D>& v, matrix::Matrix<const T, D>& mat_a,
    -
    348  bool force_copy = false) {
    -
    349  namespace ex = pika::execution::experimental;
    -
    350 
    -
    351  using pika::execution::thread_priority;
    -
    352 
    -
    353  // Note:
    -
    354  // Reflectors are stored in the lower triangular part of the A matrix leading to sharing memory
    -
    355  // between reflectors and results, which are in the upper triangular part. The problem exists only
    -
    356  // for the first tile (of the V, i.e. band excluded). Since refelectors will be used in next
    -
    357  // computations, they should be well-formed, i.e. a unit lower trapezoidal matrix. For this reason,
    -
    358  // a support tile is used, where just the reflectors values are copied, the diagonal is set to 1
    -
    359  // and the rest is zeroed out.
    -
    360  auto it_begin = panel_view.iteratorLocal().begin();
    -
    361  auto it_end = panel_view.iteratorLocal().end();
    -
    362 
    -
    363  if (has_head) {
    -
    364  const LocalTileIndex i = *it_begin;
    -
    365  matrix::SubTileSpec spec = panel_view(i);
    -
    366 
    -
    367  // Note:
    -
    368  // If the number of reflectors are limited by height (|reflector| > 1), the panel is narrower than
    -
    369  // the blocksize, leading to just using a part of A (first full nrefls columns)
    -
    370  spec.size = {spec.size.rows(), std::min(nrefls, spec.size.cols())};
    -
    371 
    -
    372  // Note:
    -
    373  // copy + laset is done in two independent tasks, but it could be theoretically merged to into a
    -
    374  // single task doing both.
    -
    375  const auto p = dlaf::internal::Policy<B>(thread_priority::high);
    -
    376  ex::start_detached(dlaf::internal::whenAllLift(splitTile(mat_a.read(i), spec), v.readwrite(i)) |
    -
    377  matrix::copy(p));
    -
    378  ex::start_detached(dlaf::internal::whenAllLift(blas::Uplo::Upper, T(0), T(1), v.readwrite(i)) |
    -
    379  tile::laset(p));
    -
    380 
    -
    381  ++it_begin;
    -
    382  }
    -
    383 
    -
    384  // The rest of the V panel of reflectors can just point to the values in A, since they are
    -
    385  // well formed in-place.
    -
    386  for (auto it = it_begin; it < it_end; ++it) {
    -
    387  const LocalTileIndex idx = *it;
    -
    388  const matrix::SubTileSpec& spec = panel_view(idx);
    -
    389 
    -
    390  // Note: This is a workaround for the deadlock problem with sub-tiles.
    -
    391  // Without this copy, during matrix update the same tile would get accessed at the same
    -
    392  // time both in readonly mode (for reflectors) and in readwrite mode (for updating the
    -
    393  // matrix). This would result in a deadlock, so instead of linking the panel to an external
    -
    394  // tile, memory provided internally by the panel is used as support. In this way, the two
    -
    395  // subtiles used in the operation belong to different tiles.
    -
    396  if (force_copy)
    -
    397  ex::start_detached(ex::when_all(matrix::splitTile(mat_a.read(idx), spec), v.readwrite(idx)) |
    -
    398  matrix::copy(dlaf::internal::Policy<B>(thread_priority::high)));
    -
    399  else
    -
    400  v.setTile(idx, matrix::splitTile(mat_a.read(idx), spec));
    -
    401  }
    -
    402 }
    -
    403 
    -
    404 template <Backend B, Device D, class T>
    -
    405 void trmmComputeW(matrix::Panel<Coord::Col, T, D>& w, matrix::Panel<Coord::Col, T, D>& v,
    -
    406  matrix::ReadOnlyTileSender<T, D> tile_t) {
    -
    407  namespace ex = pika::execution::experimental;
    -
    408 
    -
    409  using pika::execution::thread_priority;
    -
    410  using namespace blas;
    -
    411 
    -
    412  auto it = w.iteratorLocal();
    -
    413 
    -
    414  for (const auto& index_i : it) {
    -
    415  ex::start_detached(dlaf::internal::whenAllLift(Side::Right, Uplo::Upper, Op::NoTrans, Diag::NonUnit,
    -
    416  T(1), tile_t, v.read(index_i), w.readwrite(index_i)) |
    -
    417  tile::trmm3(dlaf::internal::Policy<B>(thread_priority::high)));
    -
    418  }
    -
    419 
    -
    420  if (it.empty()) {
    -
    421  ex::start_detached(std::move(tile_t));
    -
    422  }
    -
    423 }
    -
    424 
    -
    425 template <Backend B, Device D, class T>
    -
    426 void gemmUpdateX(matrix::Panel<Coord::Col, T, D>& x, matrix::Matrix<const T, D>& w2,
    -
    427  matrix::Panel<Coord::Col, const T, D>& v) {
    -
    428  namespace ex = pika::execution::experimental;
    -
    429 
    -
    430  using pika::execution::thread_priority;
    -
    431  using namespace blas;
    -
    432 
    -
    433  // GEMM X = X - 0.5 . V . W2
    -
    434  for (const auto& index_i : v.iteratorLocal())
    -
    435  ex::start_detached(dlaf::internal::whenAllLift(Op::NoTrans, Op::NoTrans, T(-0.5), v.read(index_i),
    -
    436  w2.read(LocalTileIndex(0, 0)), T(1),
    -
    437  x.readwrite(index_i)) |
    -
    438  tile::gemm(dlaf::internal::Policy<B>(thread_priority::high)));
    -
    439 }
    -
    440 
    -
    441 template <Backend B, Device D, class T>
    -
    442 void hemmComputeX(matrix::Panel<Coord::Col, T, D>& x, const matrix::SubMatrixView& view,
    -
    443  matrix::Matrix<const T, D>& a, matrix::Panel<Coord::Col, const T, D>& w) {
    -
    444  namespace ex = pika::execution::experimental;
    -
    445 
    -
    446  using pika::execution::thread_priority;
    -
    447 
    -
    448  const auto dist = a.distribution();
    -
    449 
    -
    450  // Note:
    -
    451  // They have to be set to zero, because all tiles are going to be reduced, and some tiles may not get
    -
    452  // "initialized" during computation, so they should not contribute with any spurious value to the final
    -
    453  // result.
    -
    454  matrix::util::set0<B>(thread_priority::high, x);
    -
    455 
    -
    456  const LocalTileIndex at_offset = view.begin();
    -
    457 
    -
    458  for (SizeType i = at_offset.row(); i < dist.localNrTiles().rows(); ++i) {
    -
    459  const auto limit = i + 1;
    -
    460  for (SizeType j = limit - 1; j >= at_offset.col(); --j) {
    -
    461  const LocalTileIndex ij{i, j};
    -
    462 
    -
    463  const bool is_diagonal_tile = (ij.row() == ij.col());
    -
    464 
    -
    465  const auto& tile_a = splitTile(a.read(ij), view(ij));
    -
    466 
    -
    467  if (is_diagonal_tile) {
    -
    468  hemmDiag<B>(thread_priority::high, tile_a, w.read(ij), x.readwrite(ij));
    -
    469  }
    -
    470  else {
    -
    471  // Note:
    -
    472  // Because A is hermitian and just the lower part contains the data, for each a(ij) not
    -
    473  // on the diagonal, two computations are done:
    -
    474  // - using a(ij) in its position;
    -
    475  // - using a(ij) in its "transposed" position (applying the ConjTrans to its data)
    -
    476 
    -
    477  {
    -
    478  const LocalTileIndex index_x(Coord::Row, ij.row());
    -
    479  const LocalTileIndex index_w(Coord::Row, ij.col());
    -
    480  hemmOffDiag<B>(thread_priority::high, blas::Op::NoTrans, tile_a, w.read(index_w),
    -
    481  x.readwrite(index_x));
    -
    482  }
    -
    483 
    -
    484  {
    -
    485  const LocalTileIndex index_pretended = transposed(ij);
    -
    486  const LocalTileIndex index_x(Coord::Row, index_pretended.row());
    -
    487  const LocalTileIndex index_w(Coord::Row, index_pretended.col());
    -
    488  hemmOffDiag<B>(thread_priority::high, blas::Op::ConjTrans, tile_a, w.read(index_w),
    -
    489  x.readwrite(index_x));
    -
    490  }
    -
    491  }
    -
    492  }
    -
    493  }
    -
    494 }
    -
    495 
    -
    496 template <Backend B, Device D, class T>
    -
    497 void gemmComputeW2(matrix::Matrix<T, D>& w2, matrix::Panel<Coord::Col, const T, D>& w,
    -
    498  matrix::Panel<Coord::Col, const T, D>& x) {
    -
    499  using pika::execution::thread_priority;
    -
    500 
    -
    501  namespace ex = pika::execution::experimental;
    -
    502 
    -
    503  // Note:
    -
    504  // Not all ranks in the column always hold at least a tile in the panel Ai, but all ranks in
    -
    505  // the column are going to participate to the reduce. For them, it is important to set the
    -
    506  // partial result W2 to zero.
    -
    507  ex::start_detached(w2.readwrite(LocalTileIndex(0, 0)) |
    -
    508  tile::set0(dlaf::internal::Policy<B>(thread_priority::high)));
    -
    509 
    -
    510  using namespace blas;
    -
    511  // GEMM W2 = W* . X
    -
    512  for (const auto& index_tile : w.iteratorLocal())
    -
    513  ex::start_detached(dlaf::internal::whenAllLift(Op::ConjTrans, Op::NoTrans, T(1), w.read(index_tile),
    -
    514  x.read(index_tile), T(1),
    -
    515  w2.readwrite(LocalTileIndex(0, 0))) |
    -
    516  tile::gemm(dlaf::internal::Policy<B>(thread_priority::high)));
    -
    517 }
    -
    518 
    -
    519 template <Backend B, Device D, class T>
    -
    520 void her2kUpdateTrailingMatrix(const matrix::SubMatrixView& view, matrix::Matrix<T, D>& a,
    -
    521  matrix::Panel<Coord::Col, const T, D>& x,
    -
    522  matrix::Panel<Coord::Col, const T, D>& v) {
    -
    523  static_assert(std::is_signed_v<BaseType<T>>, "alpha in computations requires to be -1");
    -
    524 
    -
    525  using pika::execution::thread_priority;
    -
    526 
    -
    527  const auto dist = a.distribution();
    -
    528 
    -
    529  const LocalTileIndex at_start = view.begin();
    -
    530 
    -
    531  for (SizeType i = at_start.row(); i < dist.localNrTiles().rows(); ++i) {
    -
    532  const auto limit = dist.template nextLocalTileFromGlobalTile<Coord::Col>(
    -
    533  dist.template globalTileFromLocalTile<Coord::Row>(i) + 1);
    -
    534  for (SizeType j = at_start.col(); j < limit; ++j) {
    -
    535  const LocalTileIndex ij_local{i, j};
    -
    536  const GlobalTileIndex ij = dist.globalTileIndex(ij_local);
    -
    537 
    -
    538  const bool is_diagonal_tile = (ij.row() == ij.col());
    -
    539 
    -
    540  auto getSubA = [&a, &view, ij_local]() {
    -
    541  return splitTile(a.readwrite(ij_local), view(ij_local));
    -
    542  };
    -
    543 
    -
    544  // The first column of the trailing matrix (except for the very first global tile) has to be
    -
    545  // updated first, in order to unlock the next iteration as soon as possible.
    -
    546  const auto priority = (j == at_start.col()) ? thread_priority::high : thread_priority::normal;
    -
    547 
    -
    548  if (is_diagonal_tile) {
    -
    549  her2kDiag<B>(priority, v.read(ij_local), x.read(ij_local), getSubA());
    -
    550  }
    -
    551  else {
    -
    552  // A -= X . V*
    -
    553  her2kOffDiag<B>(priority, x.read(ij_local), v.read(transposed(ij_local)), getSubA());
    -
    554 
    -
    555  // A -= V . X*
    -
    556  her2kOffDiag<B>(priority, v.read(ij_local), x.read(transposed(ij_local)), getSubA());
    -
    557  }
    -
    558  }
    -
    559  }
    -
    560 }
    -
    561 
    -
    562 }
    -
    563 
    -
    564 namespace distributed {
    -
    565 template <Device D, class T>
    -
    566 T computeReflector(const bool has_head, comm::Communicator& communicator,
    -
    567  const std::vector<matrix::Tile<T, D>>& panel, SizeType j) {
    -
    568  std::array<T, 2> x0_and_squares = computeX0AndSquares(has_head, panel, j);
    -
    569 
    -
    570  // Note:
    -
    571  // This is an optimization for grouping two separate low bandwidth communications, respectively
    -
    572  // bcast(x0) and reduce(norm), where the latency was degrading performances.
    -
    573  //
    -
    574  // In particular this allReduce allows to:
    -
    575  // - bcast x0, since for all ranks is 0 and just the root rank has the real value;
    -
    576  // - allReduce squares for the norm computation.
    -
    577  //
    -
    578  // Moreover, by all-reducing squares and broadcasting x0, all ranks have all the information to
    -
    579  // update locally the reflectors (section they have). This is more efficient than computing params
    -
    580  // (e.g. norm, y, tau) just on the root rank and then having to broadcast them (i.e. additional
    -
    581  // communication).
    -
    582  comm::sync::allReduceInPlace(communicator, MPI_SUM,
    -
    583  common::make_data(x0_and_squares.data(),
    -
    584  to_SizeType(x0_and_squares.size())));
    -
    585 
    -
    586  auto tau = computeReflectorAndTau(has_head, panel, j, std::move(x0_and_squares));
    -
    587 
    -
    588  return tau;
    -
    589 }
    -
    590 
    -
    591 template <class MatrixLikeA, class MatrixLikeTaus, class TriggerSender, class CommSender>
    -
    592 void computePanelReflectors(TriggerSender&& trigger, comm::IndexT_MPI rank_v0,
    -
    593  CommSender&& mpi_col_chain_panel, MatrixLikeA& mat_a,
    -
    594  MatrixLikeTaus& mat_taus, SizeType j_sub,
    -
    595  const matrix::SubPanelView& panel_view) {
    -
    596  static Device constexpr D = MatrixLikeA::device;
    -
    597  using T = typename MatrixLikeA::ElementType;
    -
    598  namespace ex = pika::execution::experimental;
    -
    599  namespace di = dlaf::internal;
    -
    600 
    -
    601  std::vector<matrix::ReadWriteTileSender<T, D>> panel_tiles;
    -
    602  panel_tiles.reserve(to_sizet(std::distance(panel_view.iteratorLocal().begin(),
    -
    603  panel_view.iteratorLocal().end())));
    -
    604  for (const auto& i : panel_view.iteratorLocal()) {
    -
    605  const matrix::SubTileSpec& spec = panel_view(i);
    -
    606  panel_tiles.emplace_back(matrix::splitTile(mat_a.readwrite(i), spec));
    -
    607  }
    -
    608 
    -
    609  const std::size_t nthreads = getReductionToBandPanelNWorkers();
    -
    610  auto s =
    -
    611  ex::when_all(ex::just(std::make_unique<pika::barrier<>>(nthreads),
    -
    612  std::vector<common::internal::vector<T>>{}), // w (internally required)
    -
    613  mat_taus.readwrite(GlobalTileIndex(j_sub, 0)),
    -
    614  ex::when_all_vector(std::move(panel_tiles)),
    -
    615  std::forward<CommSender>(mpi_col_chain_panel), std::forward<TriggerSender>(trigger)) |
    -
    616  ex::transfer(di::getBackendScheduler<Backend::MC>(pika::execution::thread_priority::high)) |
    -
    617  ex::bulk(nthreads, [nthreads, rank_v0,
    -
    618  cols = panel_view.cols()](const std::size_t index, auto& barrier_ptr, auto& w,
    -
    619  auto& taus, auto& tiles, auto&& pcomm) {
    -
    620  const bool rankHasHead = rank_v0 == pcomm.get().rank();
    -
    621 
    -
    622  const auto barrier_busy_wait = getReductionToBandBarrierBusyWait();
    -
    623  const std::size_t batch_size = util::ceilDiv(tiles.size(), nthreads);
    -
    624  const std::size_t begin = index * batch_size;
    -
    625  const std::size_t end = std::min(index * batch_size + batch_size, tiles.size());
    -
    626  const SizeType nrefls = taus.size().rows();
    -
    627 
    -
    628  if (index == 0) {
    -
    629  w.resize(nthreads);
    -
    630  }
    -
    631 
    -
    632  for (SizeType j = 0; j < nrefls; ++j) {
    -
    633  // STEP1: compute tau and reflector (single-thread)
    -
    634  if (index == 0) {
    -
    635  const bool has_head = rankHasHead;
    -
    636  taus({j, 0}) = computeReflector(has_head, pcomm.get(), tiles, j);
    -
    637  }
    -
    638  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    -
    639 
    -
    640  // STEP2a: compute w (multi-threaded)
    -
    641  const SizeType pt_cols = cols - (j + 1);
    -
    642  if (pt_cols == 0)
    -
    643  break;
    -
    644 
    -
    645  const bool has_head = rankHasHead && (index == 0);
    -
    646 
    -
    647  w[index] = common::internal::vector<T>(pt_cols, 0);
    -
    648  computeWTrailingPanel(has_head, tiles, w[index], j, pt_cols, begin, end);
    -
    649  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    -
    650 
    -
    651  // STEP2b: reduce w results (single-threaded)
    -
    652  if (index == 0) {
    -
    653  dlaf::eigensolver::internal::reduceColumnVectors(w);
    -
    654  comm::sync::allReduceInPlace(pcomm.get(), MPI_SUM, common::make_data(w[0].data(), pt_cols));
    -
    655  }
    -
    656  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    -
    657 
    -
    658  // STEP3: update trailing panel (multi-threaded)
    -
    659  updateTrailingPanel(has_head, tiles, j, w[0], taus({j, 0}), begin, end);
    -
    660  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    -
    661  }
    -
    662  });
    -
    663  ex::start_detached(std::move(s));
    -
    664 }
    -
    665 
    -
    666 template <Backend B, Device D, class T>
    -
    667 void hemmComputeX(comm::IndexT_MPI reducer_col, matrix::Panel<Coord::Col, T, D>& x,
    -
    668  matrix::Panel<Coord::Row, T, D, matrix::StoreTransposed::Yes>& xt,
    -
    669  const matrix::SubMatrixView& view, matrix::Matrix<const T, D>& a,
    -
    670  matrix::Panel<Coord::Col, const T, D>& w,
    -
    671  matrix::Panel<Coord::Row, const T, D, matrix::StoreTransposed::Yes>& wt,
    -
    672  common::Pipeline<comm::Communicator>& mpi_row_chain,
    -
    673  common::Pipeline<comm::Communicator>& mpi_col_chain) {
    -
    674  namespace ex = pika::execution::experimental;
    -
    675 
    -
    676  using pika::execution::thread_priority;
    -
    677 
    -
    678  const auto dist = a.distribution();
    -
    679  const auto rank = dist.rankIndex();
    -
    680 
    -
    681  // Note:
    -
    682  // They have to be set to zero, because all tiles are going to be reduced, and some tiles may not get
    -
    683  // "initialized" during computation, so they should not contribute with any spurious value to the final
    -
    684  // result.
    -
    685  matrix::util::set0<B>(thread_priority::high, x);
    -
    686  matrix::util::set0<B>(thread_priority::high, xt);
    -
    687 
    -
    688  const LocalTileIndex at_offset = view.begin();
    -
    689 
    -
    690  for (SizeType i = at_offset.row(); i < dist.localNrTiles().rows(); ++i) {
    -
    691  const auto limit = dist.template nextLocalTileFromGlobalTile<Coord::Col>(
    -
    692  dist.template globalTileFromLocalTile<Coord::Row>(i) + 1);
    -
    693  for (SizeType j = limit - 1; j >= at_offset.col(); --j) {
    -
    694  const LocalTileIndex ij_local{i, j};
    -
    695  const GlobalTileIndex ij = dist.globalTileIndex(ij_local);
    -
    696 
    -
    697  const bool is_diagonal_tile = (ij.row() == ij.col());
    -
    698 
    -
    699  auto tile_a = splitTile(a.read(ij), view(ij_local));
    -
    700 
    -
    701  if (is_diagonal_tile) {
    -
    702  hemmDiag<B>(thread_priority::high, std::move(tile_a), w.read(ij_local), x.readwrite(ij_local));
    -
    703  }
    -
    704  else {
    -
    705  // Note:
    -
    706  // Since it is not a diagonal tile, otherwise it would have been managed in the previous
    -
    707  // branch, the second operand is not available in W but it is accessible through the
    -
    708  // support panel Wt.
    -
    709  // However, since we are still computing the "straight" part, the result can be stored
    -
    710  // in the "local" panel X.
    -
    711  hemmOffDiag<B>(thread_priority::high, blas::Op::NoTrans, tile_a, wt.read(ij_local),
    -
    712  x.readwrite(ij_local));
    -
    713 
    -
    714  // Note:
    -
    715  // Here we are considering the hermitian part of A, so coordinates have to be "mirrored".
    -
    716  // So, first step is identifying the mirrored cell coordinate, i.e. swap row/col, together
    -
    717  // with realizing if the new coord lays on an owned row or not.
    -
    718  // If yes, the result can be stored in the X, otherwise Xt support panel will be used.
    -
    719  // For what concerns the second operand, it can be found for sure in W. In fact, the
    -
    720  // multiplication requires matching col(A) == row(W), but since coordinates are mirrored,
    -
    721  // we are matching row(A) == row(W), so it is local by construction.
    -
    722  const auto owner = dist.template rankGlobalTile<Coord::Row>(ij.col());
    -
    723 
    -
    724  const LocalTileIndex index_x{dist.template localTileFromGlobalTile<Coord::Row>(ij.col()), 0};
    -
    725  const LocalTileIndex index_xt{0, ij_local.col()};
    -
    726 
    -
    727  auto tile_x = (dist.rankIndex().row() == owner) ? x.readwrite(index_x) : xt.readwrite(index_xt);
    -
    728 
    -
    729  hemmOffDiag<B>(thread_priority::high, blas::Op::ConjTrans, std::move(tile_a), w.read(ij_local),
    -
    730  std::move(tile_x));
    -
    731  }
    -
    732  }
    -
    733  }
    -
    734 
    -
    735  // Note:
    -
    736  // At this point, partial results of X and Xt are available in the panels, and they have to be reduced,
    -
    737  // both row-wise and col-wise.
    -
    738  // The final X result will be available just on Ai panel column.
    -
    739 
    -
    740  // Note:
    -
    741  // The first step in reducing partial results distributed over X and Xt, it is to reduce the row
    -
    742  // panel Xt col-wise, by collecting all Xt results on the rank which can "mirror" the result on its
    -
    743  // rows (i.e. diagonal). So, for each tile of the row panel, select who is the "diagonal" rank that can
    -
    744  // mirror and reduce on it.
    -
    745  for (const auto& index_xt : xt.iteratorLocal()) {
    -
    746  const auto index_k = dist.template globalTileFromLocalTile<Coord::Col>(index_xt.col());
    -
    747  const auto rank_owner_row = dist.template rankGlobalTile<Coord::Row>(index_k);
    -
    748 
    -
    749  if (rank_owner_row == rank.row()) {
    -
    750  // Note:
    -
    751  // Since it is the owner, it has to perform the "mirroring" of the results from columns to
    -
    752  // rows.
    -
    753  //
    -
    754  // Moreover, it reduces in place because the owner of the diagonal stores the partial result
    -
    755  // directly in x (without using xt)
    -
    756  const auto i = dist.template localTileFromGlobalTile<Coord::Row>(index_k);
    -
    757  ex::start_detached(comm::scheduleReduceRecvInPlace(mpi_col_chain(), MPI_SUM, x.readwrite({i, 0})));
    -
    758  }
    -
    759  else {
    -
    760  ex::start_detached(comm::scheduleReduceSend(mpi_col_chain(), rank_owner_row, MPI_SUM,
    -
    761  xt.read(index_xt)));
    -
    762  }
    -
    763  }
    -
    764 
    -
    765  // Note:
    -
    766  // At this point partial results are all collected in X (Xt has been embedded in previous step),
    -
    767  // so the last step needed is to reduce these last partial results in the final results.
    -
    768  // The result is needed just on the column with reflectors.
    -
    769  for (const auto& index_x : x.iteratorLocal()) {
    -
    770  if (reducer_col == rank.col())
    -
    771  ex::start_detached(comm::scheduleReduceRecvInPlace(mpi_row_chain(), MPI_SUM,
    -
    772  x.readwrite(index_x)));
    -
    773  else
    -
    774  ex::start_detached(comm::scheduleReduceSend(mpi_row_chain(), reducer_col, MPI_SUM,
    -
    775  x.read(index_x)));
    -
    776  }
    -
    777 }
    -
    778 
    -
    779 template <Backend B, Device D, class T>
    -
    780 void her2kUpdateTrailingMatrix(const matrix::SubMatrixView& view, Matrix<T, D>& a,
    -
    781  matrix::Panel<Coord::Col, const T, D>& x,
    -
    782  matrix::Panel<Coord::Row, const T, D, matrix::StoreTransposed::Yes>& vt,
    -
    783  matrix::Panel<Coord::Col, const T, D>& v,
    -
    784  matrix::Panel<Coord::Row, const T, D, matrix::StoreTransposed::Yes>& xt) {
    -
    785  static_assert(std::is_signed_v<BaseType<T>>, "alpha in computations requires to be -1");
    -
    786 
    -
    787  using pika::execution::thread_priority;
    -
    788 
    -
    789  const auto dist = a.distribution();
    -
    790 
    -
    791  const LocalTileIndex at_start = view.begin();
    -
    792 
    -
    793  for (SizeType i = at_start.row(); i < dist.localNrTiles().rows(); ++i) {
    -
    794  const auto limit = dist.template nextLocalTileFromGlobalTile<Coord::Col>(
    -
    795  dist.template globalTileFromLocalTile<Coord::Row>(i) + 1);
    -
    796  for (SizeType j = at_start.col(); j < limit; ++j) {
    -
    797  const LocalTileIndex ij_local{i, j};
    -
    798  const GlobalTileIndex ij = dist.globalTileIndex(ij_local);
    -
    799 
    -
    800  const bool is_diagonal_tile = (ij.row() == ij.col());
    -
    801 
    -
    802  auto getSubA = [&a, &view, ij_local]() {
    -
    803  return splitTile(a.readwrite(ij_local), view(ij_local));
    -
    804  };
    -
    805 
    -
    806  // The first column of the trailing matrix (except for the very first global tile) has to be
    -
    807  // updated first, in order to unlock the next iteration as soon as possible.
    -
    808  const auto priority = (j == at_start.col()) ? thread_priority::high : thread_priority::normal;
    -
    809 
    -
    810  if (is_diagonal_tile) {
    -
    811  her2kDiag<B>(priority, v.read(ij_local), x.read(ij_local), getSubA());
    -
    812  }
    -
    813  else {
    -
    814  // A -= X . V*
    -
    815  her2kOffDiag<B>(priority, x.read(ij_local), vt.read(ij_local), getSubA());
    -
    816 
    -
    817  // A -= V . X*
    -
    818  her2kOffDiag<B>(priority, v.read(ij_local), xt.read(ij_local), getSubA());
    -
    819  }
    -
    820  }
    -
    821  }
    +
    45 #include <dlaf/matrix/tile.h>
    +
    46 #include <dlaf/matrix/views.h>
    +
    47 #include <dlaf/schedulers.h>
    +
    48 #include <dlaf/sender/traits.h>
    +
    49 #include <dlaf/types.h>
    +
    50 #include <dlaf/util_math.h>
    +
    51 #include <dlaf/util_matrix.h>
    +
    52 
    +
    53 namespace dlaf::eigensolver::internal {
    +
    54 
    +
    55 // Given a vector of vectors, reduce all vectors in the first one using sum operation
    +
    56 template <class T>
    +
    57 void reduceColumnVectors(std::vector<common::internal::vector<T>>& columnVectors) {
    +
    58  for (std::size_t i = 1; i < columnVectors.size(); ++i) {
    +
    59  DLAF_ASSERT_HEAVY(columnVectors[0].size() == columnVectors[i].size(), columnVectors[0].size(),
    +
    60  columnVectors[i].size());
    +
    61  for (SizeType j = 0; j < columnVectors[0].size(); ++j)
    +
    62  columnVectors[0][j] += columnVectors[i][j];
    +
    63  }
    +
    64 }
    +
    65 
    +
    66 namespace red2band {
    +
    67 
    +
    68 // Extract x0 and compute local cumulative sum of squares of the reflector column
    +
    69 template <Device D, class T>
    +
    70 std::array<T, 2> computeX0AndSquares(const bool has_head, const std::vector<matrix::Tile<T, D>>& panel,
    +
    71  SizeType j) {
    +
    72  std::array<T, 2> x0_and_squares{0, 0};
    +
    73  auto it_begin = panel.begin();
    +
    74  auto it_end = panel.end();
    +
    75 
    +
    76  common::internal::SingleThreadedBlasScope single;
    +
    77 
    +
    78  if (has_head) {
    +
    79  auto& tile_v0 = *it_begin++;
    +
    80 
    +
    81  const TileElementIndex idx_x0(j, j);
    +
    82  x0_and_squares[0] = tile_v0(idx_x0);
    +
    83 
    +
    84  T* reflector_ptr = tile_v0.ptr({idx_x0});
    +
    85  x0_and_squares[1] =
    +
    86  blas::dot(tile_v0.size().rows() - idx_x0.row(), reflector_ptr, 1, reflector_ptr, 1);
    +
    87  }
    +
    88 
    +
    89  for (auto it = it_begin; it != it_end; ++it) {
    +
    90  const auto& tile = *it;
    +
    91 
    +
    92  T* reflector_ptr = tile.ptr({0, j});
    +
    93  x0_and_squares[1] += blas::dot(tile.size().rows(), reflector_ptr, 1, reflector_ptr, 1);
    +
    94  }
    +
    95  return x0_and_squares;
    +
    96 }
    +
    97 
    +
    98 template <Device D, class T>
    +
    99 T computeReflectorAndTau(const bool has_head, const std::vector<matrix::Tile<T, D>>& panel,
    +
    100  const SizeType j, std::array<T, 2> x0_and_squares) {
    +
    101  const T norm = std::sqrt(x0_and_squares[1]);
    +
    102  const T x0 = x0_and_squares[0];
    +
    103  const T y = std::signbit(std::real(x0_and_squares[0])) ? norm : -norm;
    +
    104  const T tau = (y - x0) / y;
    +
    105 
    +
    106  auto it_begin = panel.begin();
    +
    107  auto it_end = panel.end();
    +
    108 
    +
    109  common::internal::SingleThreadedBlasScope single;
    +
    110 
    +
    111  if (has_head) {
    +
    112  const auto& tile_v0 = *it_begin++;
    +
    113 
    +
    114  const TileElementIndex idx_x0(j, j);
    +
    115  tile_v0(idx_x0) = y;
    +
    116 
    +
    117  if (j + 1 < tile_v0.size().rows()) {
    +
    118  T* v = tile_v0.ptr({j + 1, j});
    +
    119  blas::scal(tile_v0.size().rows() - (j + 1), T(1) / (x0 - y), v, 1);
    +
    120  }
    +
    121  }
    +
    122 
    +
    123  for (auto it = it_begin; it != it_end; ++it) {
    +
    124  auto& tile_v = *it;
    +
    125  T* v = tile_v.ptr({0, j});
    +
    126  blas::scal(tile_v.size().rows(), T(1) / (x0 - y), v, 1);
    +
    127  }
    +
    128 
    +
    129  return tau;
    +
    130 }
    +
    131 
    +
    132 template <Device D, class T>
    +
    133 void computeWTrailingPanel(const bool has_head, const std::vector<matrix::Tile<T, D>>& panel,
    +
    134  common::internal::vector<T>& w, SizeType j, const SizeType pt_cols,
    +
    135  const std::size_t begin, const std::size_t end) {
    +
    136  // for each tile in the panel, consider just the trailing panel
    +
    137  // i.e. all rows (height = reflector), just columns to the right of the current reflector
    +
    138  if (!(pt_cols > 0))
    +
    139  return;
    +
    140 
    +
    141  const TileElementIndex index_el_x0(j, j);
    +
    142  bool has_first_component = has_head;
    +
    143 
    +
    144  common::internal::SingleThreadedBlasScope single;
    +
    145 
    +
    146  // W = Pt* . V
    +
    147  for (auto index = begin; index < end; ++index) {
    +
    148  const matrix::Tile<const T, D>& tile_a = panel[index];
    +
    149  const SizeType first_element = has_first_component ? index_el_x0.row() : 0;
    +
    150 
    +
    151  TileElementIndex pt_start{first_element, index_el_x0.col() + 1};
    +
    152  TileElementSize pt_size{tile_a.size().rows() - pt_start.row(), pt_cols};
    +
    153  TileElementIndex v_start{first_element, index_el_x0.col()};
    +
    154 
    +
    155  if (has_first_component) {
    +
    156  const TileElementSize offset{1, 0};
    +
    157 
    +
    158  const T fake_v = 1;
    +
    159  blas::gemv(blas::Layout::ColMajor, blas::Op::ConjTrans, offset.rows(), pt_size.cols(), T(1),
    +
    160  tile_a.ptr(pt_start), tile_a.ld(), &fake_v, 1, T(0), w.data(), 1);
    +
    161 
    +
    162  pt_start = pt_start + offset;
    +
    163  v_start = v_start + offset;
    +
    164  pt_size = pt_size - offset;
    +
    165 
    +
    166  has_first_component = false;
    +
    167  }
    +
    168 
    +
    169  if (pt_start.isIn(tile_a.size())) {
    +
    170  // W += 1 . A* . V
    +
    171  blas::gemv(blas::Layout::ColMajor, blas::Op::ConjTrans, pt_size.rows(), pt_size.cols(), T(1),
    +
    172  tile_a.ptr(pt_start), tile_a.ld(), tile_a.ptr(v_start), 1, T(1), w.data(), 1);
    +
    173  }
    +
    174  }
    +
    175 }
    +
    176 
    +
    177 template <Device D, class T>
    +
    178 void updateTrailingPanel(const bool has_head, const std::vector<matrix::Tile<T, D>>& panel, SizeType j,
    +
    179  const std::vector<T>& w, const T tau, const std::size_t begin,
    +
    180  const std::size_t end) {
    +
    181  const TileElementIndex index_el_x0(j, j);
    +
    182 
    +
    183  bool has_first_component = has_head;
    +
    184 
    +
    185  common::internal::SingleThreadedBlasScope single;
    +
    186 
    +
    187  // GER Pt = Pt - tau . v . w*
    +
    188  for (auto index = begin; index < end; ++index) {
    +
    189  const matrix::Tile<T, D>& tile_a = panel[index];
    +
    190  const SizeType first_element = has_first_component ? index_el_x0.row() : 0;
    +
    191 
    +
    192  TileElementIndex pt_start{first_element, index_el_x0.col() + 1};
    +
    193  TileElementSize pt_size{tile_a.size().rows() - pt_start.row(),
    +
    194  tile_a.size().cols() - pt_start.col()};
    +
    195  TileElementIndex v_start{first_element, index_el_x0.col()};
    +
    196 
    +
    197  if (has_first_component) {
    +
    198  const TileElementSize offset{1, 0};
    +
    199 
    +
    200  // Pt = Pt - tau * v[0] * w*
    +
    201  const T fake_v = 1;
    +
    202  blas::ger(blas::Layout::ColMajor, 1, pt_size.cols(), -dlaf::conj(tau), &fake_v, 1, w.data(), 1,
    +
    203  tile_a.ptr(pt_start), tile_a.ld());
    +
    204 
    +
    205  pt_start = pt_start + offset;
    +
    206  v_start = v_start + offset;
    +
    207  pt_size = pt_size - offset;
    +
    208 
    +
    209  has_first_component = false;
    +
    210  }
    +
    211 
    +
    212  if (pt_start.isIn(tile_a.size())) {
    +
    213  // Pt = Pt - tau * v * w*
    +
    214  blas::ger(blas::Layout::ColMajor, pt_size.rows(), pt_size.cols(), -dlaf::conj(tau),
    +
    215  tile_a.ptr(v_start), 1, w.data(), 1, tile_a.ptr(pt_start), tile_a.ld());
    +
    216  }
    +
    217  }
    +
    218 }
    +
    219 
    +
    220 template <Backend B, typename ASender, typename WSender, typename XSender>
    +
    221 void hemmDiag(pika::execution::thread_priority priority, ASender&& tile_a, WSender&& tile_w,
    +
    222  XSender&& tile_x) {
    +
    223  using T = dlaf::internal::SenderElementType<ASender>;
    +
    224  pika::execution::experimental::start_detached(
    +
    225  dlaf::internal::whenAllLift(blas::Side::Left, blas::Uplo::Lower, T(1),
    +
    226  std::forward<ASender>(tile_a), std::forward<WSender>(tile_w), T(1),
    +
    227  std::forward<XSender>(tile_x)) |
    +
    228  tile::hemm(dlaf::internal::Policy<B>(priority)));
    +
    229 }
    +
    230 
    +
    231 // X += op(A) * W
    +
    232 template <Backend B, typename ASender, typename WSender, typename XSender>
    +
    233 void hemmOffDiag(pika::execution::thread_priority priority, blas::Op op, ASender&& tile_a,
    +
    234  WSender&& tile_w, XSender&& tile_x) {
    +
    235  using T = dlaf::internal::SenderElementType<ASender>;
    +
    236  pika::execution::experimental::start_detached(
    +
    237  dlaf::internal::whenAllLift(op, blas::Op::NoTrans, T(1), std::forward<ASender>(tile_a),
    +
    238  std::forward<WSender>(tile_w), T(1), std::forward<XSender>(tile_x)) |
    +
    239  tile::gemm(dlaf::internal::Policy<B>(priority)));
    +
    240 }
    +
    241 
    +
    242 template <Backend B, typename VSender, typename XSender, typename ASender>
    +
    243 void her2kDiag(pika::execution::thread_priority priority, VSender&& tile_v, XSender&& tile_x,
    +
    244  ASender&& tile_a) {
    +
    245  using T = dlaf::internal::SenderElementType<VSender>;
    +
    246  pika::execution::experimental::start_detached(
    +
    247  dlaf::internal::whenAllLift(blas::Uplo::Lower, blas::Op::NoTrans, T(-1),
    +
    248  std::forward<VSender>(tile_v), std::forward<XSender>(tile_x),
    +
    249  BaseType<T>(1), std::forward<ASender>(tile_a)) |
    +
    250  tile::her2k(dlaf::internal::Policy<B>(priority)));
    +
    251 }
    +
    252 
    +
    253 // C -= A . B*
    +
    254 template <Backend B, typename ASender, typename BSender, typename CSender>
    +
    255 void her2kOffDiag(pika::execution::thread_priority priority, ASender&& tile_a, BSender&& tile_b,
    +
    256  CSender&& tile_c) {
    +
    257  using T = dlaf::internal::SenderElementType<ASender>;
    +
    258  pika::execution::experimental::start_detached(
    +
    259  dlaf::internal::whenAllLift(blas::Op::NoTrans, blas::Op::ConjTrans, T(-1),
    +
    260  std::forward<ASender>(tile_a), std::forward<BSender>(tile_b), T(1),
    +
    261  std::forward<CSender>(tile_c)) |
    +
    262  tile::gemm(dlaf::internal::Policy<B>(priority)));
    +
    263 }
    +
    264 
    +
    265 namespace local {
    +
    266 
    +
    267 template <Device D, class T>
    +
    268 T computeReflector(const std::vector<matrix::Tile<T, D>>& panel, SizeType j) {
    +
    269  constexpr bool has_head = true;
    +
    270 
    +
    271  std::array<T, 2> x0_and_squares = computeX0AndSquares(has_head, panel, j);
    +
    272 
    +
    273  auto tau = computeReflectorAndTau(has_head, panel, j, std::move(x0_and_squares));
    +
    274 
    +
    275  return tau;
    +
    276 }
    +
    277 
    +
    278 template <class MatrixLikeA, class MatrixLikeTaus>
    +
    279 void computePanelReflectors(MatrixLikeA& mat_a, MatrixLikeTaus& mat_taus, const SizeType j_sub,
    +
    280  const matrix::SubPanelView& panel_view) {
    +
    281  static Device constexpr D = MatrixLikeA::device;
    +
    282  using T = typename MatrixLikeA::ElementType;
    +
    283  namespace ex = pika::execution::experimental;
    +
    284  namespace di = dlaf::internal;
    +
    285 
    +
    286  std::vector<matrix::ReadWriteTileSender<T, D>> panel_tiles;
    +
    287  panel_tiles.reserve(to_sizet(std::distance(panel_view.iteratorLocal().begin(),
    +
    288  panel_view.iteratorLocal().end())));
    +
    289  for (const auto& i : panel_view.iteratorLocal()) {
    +
    290  const matrix::SubTileSpec& spec = panel_view(i);
    +
    291  panel_tiles.emplace_back(matrix::splitTile(mat_a.readwrite(i), spec));
    +
    292  }
    +
    293 
    +
    294  const std::size_t nthreads = getReductionToBandPanelNWorkers();
    +
    295  auto s =
    +
    296  ex::when_all(ex::just(std::make_unique<pika::barrier<>>(nthreads),
    +
    297  std::vector<common::internal::vector<T>>{}), // w (internally required)
    +
    298  mat_taus.readwrite(LocalTileIndex(j_sub, 0)),
    +
    299  ex::when_all_vector(std::move(panel_tiles))) |
    +
    300  ex::transfer(di::getBackendScheduler<Backend::MC>(pika::execution::thread_priority::high)) |
    +
    301  ex::bulk(nthreads, [nthreads, cols = panel_view.cols()](const std::size_t index, auto& barrier_ptr,
    +
    302  auto& w, auto& taus, auto& tiles) {
    +
    303  const auto barrier_busy_wait = getReductionToBandBarrierBusyWait();
    +
    304  const std::size_t batch_size = util::ceilDiv(tiles.size(), nthreads);
    +
    305  const std::size_t begin = index * batch_size;
    +
    306  const std::size_t end = std::min(index * batch_size + batch_size, tiles.size());
    +
    307  const SizeType nrefls = taus.size().rows();
    +
    308 
    +
    309  if (index == 0) {
    +
    310  w.resize(nthreads);
    +
    311  }
    +
    312 
    +
    313  for (SizeType j = 0; j < nrefls; ++j) {
    +
    314  // STEP1: compute tau and reflector (single-thread)
    +
    315  if (index == 0) {
    +
    316  taus({j, 0}) = computeReflector(tiles, j);
    +
    317  }
    +
    318 
    +
    319  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    +
    320 
    +
    321  // STEP2a: compute w (multi-threaded)
    +
    322  const SizeType pt_cols = cols - (j + 1);
    +
    323  if (pt_cols == 0)
    +
    324  break;
    +
    325  const bool has_head = (index == 0);
    +
    326 
    +
    327  w[index] = common::internal::vector<T>(pt_cols, 0);
    +
    328  computeWTrailingPanel(has_head, tiles, w[index], j, pt_cols, begin, end);
    +
    329  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    +
    330 
    +
    331  // STEP2b: reduce w results (single-threaded)
    +
    332  if (index == 0)
    +
    333  dlaf::eigensolver::internal::reduceColumnVectors(w);
    +
    334  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    +
    335 
    +
    336  // STEP3: update trailing panel (multi-threaded)
    +
    337  updateTrailingPanel(has_head, tiles, j, w[0], taus({j, 0}), begin, end);
    +
    338  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    +
    339  }
    +
    340  });
    +
    341  ex::start_detached(std::move(s));
    +
    342 }
    +
    343 
    +
    344 template <Backend B, Device D, class T>
    +
    345 void setupReflectorPanelV(bool has_head, const matrix::SubPanelView& panel_view, const SizeType nrefls,
    +
    346  matrix::Panel<Coord::Col, T, D>& v, matrix::Matrix<const T, D>& mat_a,
    +
    347  bool force_copy = false) {
    +
    348  namespace ex = pika::execution::experimental;
    +
    349 
    +
    350  using pika::execution::thread_priority;
    +
    351 
    +
    352  // Note:
    +
    353  // Reflectors are stored in the lower triangular part of the A matrix leading to sharing memory
    +
    354  // between reflectors and results, which are in the upper triangular part. The problem exists only
    +
    355  // for the first tile (of the V, i.e. band excluded). Since refelectors will be used in next
    +
    356  // computations, they should be well-formed, i.e. a unit lower trapezoidal matrix. For this reason,
    +
    357  // a support tile is used, where just the reflectors values are copied, the diagonal is set to 1
    +
    358  // and the rest is zeroed out.
    +
    359  auto it_begin = panel_view.iteratorLocal().begin();
    +
    360  auto it_end = panel_view.iteratorLocal().end();
    +
    361 
    +
    362  if (has_head) {
    +
    363  const LocalTileIndex i = *it_begin;
    +
    364  matrix::SubTileSpec spec = panel_view(i);
    +
    365 
    +
    366  // Note:
    +
    367  // If the number of reflectors are limited by height (|reflector| > 1), the panel is narrower than
    +
    368  // the blocksize, leading to just using a part of A (first full nrefls columns)
    +
    369  spec.size = {spec.size.rows(), std::min(nrefls, spec.size.cols())};
    +
    370 
    +
    371  // Note:
    +
    372  // copy + laset is done in two independent tasks, but it could be theoretically merged to into a
    +
    373  // single task doing both.
    +
    374  const auto p = dlaf::internal::Policy<B>(thread_priority::high);
    +
    375  ex::start_detached(dlaf::internal::whenAllLift(splitTile(mat_a.read(i), spec), v.readwrite(i)) |
    +
    376  matrix::copy(p));
    +
    377  ex::start_detached(dlaf::internal::whenAllLift(blas::Uplo::Upper, T(0), T(1), v.readwrite(i)) |
    +
    378  tile::laset(p));
    +
    379 
    +
    380  ++it_begin;
    +
    381  }
    +
    382 
    +
    383  // The rest of the V panel of reflectors can just point to the values in A, since they are
    +
    384  // well formed in-place.
    +
    385  for (auto it = it_begin; it < it_end; ++it) {
    +
    386  const LocalTileIndex idx = *it;
    +
    387  const matrix::SubTileSpec& spec = panel_view(idx);
    +
    388 
    +
    389  // Note: This is a workaround for the deadlock problem with sub-tiles.
    +
    390  // Without this copy, during matrix update the same tile would get accessed at the same
    +
    391  // time both in readonly mode (for reflectors) and in readwrite mode (for updating the
    +
    392  // matrix). This would result in a deadlock, so instead of linking the panel to an external
    +
    393  // tile, memory provided internally by the panel is used as support. In this way, the two
    +
    394  // subtiles used in the operation belong to different tiles.
    +
    395  if (force_copy)
    +
    396  ex::start_detached(ex::when_all(matrix::splitTile(mat_a.read(idx), spec), v.readwrite(idx)) |
    +
    397  matrix::copy(dlaf::internal::Policy<B>(thread_priority::high)));
    +
    398  else
    +
    399  v.setTile(idx, matrix::splitTile(mat_a.read(idx), spec));
    +
    400  }
    +
    401 }
    +
    402 
    +
    403 template <Backend B, Device D, class T>
    +
    404 void trmmComputeW(matrix::Panel<Coord::Col, T, D>& w, matrix::Panel<Coord::Col, T, D>& v,
    +
    405  matrix::ReadOnlyTileSender<T, D> tile_t) {
    +
    406  namespace ex = pika::execution::experimental;
    +
    407 
    +
    408  using pika::execution::thread_priority;
    +
    409  using namespace blas;
    +
    410 
    +
    411  auto it = w.iteratorLocal();
    +
    412 
    +
    413  for (const auto& index_i : it) {
    +
    414  ex::start_detached(dlaf::internal::whenAllLift(Side::Right, Uplo::Upper, Op::NoTrans, Diag::NonUnit,
    +
    415  T(1), tile_t, v.read(index_i), w.readwrite(index_i)) |
    +
    416  tile::trmm3(dlaf::internal::Policy<B>(thread_priority::high)));
    +
    417  }
    +
    418 
    +
    419  if (it.empty()) {
    +
    420  ex::start_detached(std::move(tile_t));
    +
    421  }
    +
    422 }
    +
    423 
    +
    424 template <Backend B, Device D, class T>
    +
    425 void gemmUpdateX(matrix::Panel<Coord::Col, T, D>& x, matrix::Matrix<const T, D>& w2,
    +
    426  matrix::Panel<Coord::Col, const T, D>& v) {
    +
    427  namespace ex = pika::execution::experimental;
    +
    428 
    +
    429  using pika::execution::thread_priority;
    +
    430  using namespace blas;
    +
    431 
    +
    432  // GEMM X = X - 0.5 . V . W2
    +
    433  for (const auto& index_i : v.iteratorLocal())
    +
    434  ex::start_detached(dlaf::internal::whenAllLift(Op::NoTrans, Op::NoTrans, T(-0.5), v.read(index_i),
    +
    435  w2.read(LocalTileIndex(0, 0)), T(1),
    +
    436  x.readwrite(index_i)) |
    +
    437  tile::gemm(dlaf::internal::Policy<B>(thread_priority::high)));
    +
    438 }
    +
    439 
    +
    440 template <Backend B, Device D, class T>
    +
    441 void hemmComputeX(matrix::Panel<Coord::Col, T, D>& x, const matrix::SubMatrixView& view,
    +
    442  matrix::Matrix<const T, D>& a, matrix::Panel<Coord::Col, const T, D>& w) {
    +
    443  namespace ex = pika::execution::experimental;
    +
    444 
    +
    445  using pika::execution::thread_priority;
    +
    446 
    +
    447  const auto dist = a.distribution();
    +
    448 
    +
    449  // Note:
    +
    450  // They have to be set to zero, because all tiles are going to be reduced, and some tiles may not get
    +
    451  // "initialized" during computation, so they should not contribute with any spurious value to the final
    +
    452  // result.
    +
    453  matrix::util::set0<B>(thread_priority::high, x);
    +
    454 
    +
    455  const LocalTileIndex at_offset = view.begin();
    +
    456 
    +
    457  for (SizeType i = at_offset.row(); i < dist.localNrTiles().rows(); ++i) {
    +
    458  const auto limit = i + 1;
    +
    459  for (SizeType j = limit - 1; j >= at_offset.col(); --j) {
    +
    460  const LocalTileIndex ij{i, j};
    +
    461 
    +
    462  const bool is_diagonal_tile = (ij.row() == ij.col());
    +
    463 
    +
    464  const auto& tile_a = splitTile(a.read(ij), view(ij));
    +
    465 
    +
    466  if (is_diagonal_tile) {
    +
    467  hemmDiag<B>(thread_priority::high, tile_a, w.read(ij), x.readwrite(ij));
    +
    468  }
    +
    469  else {
    +
    470  // Note:
    +
    471  // Because A is hermitian and just the lower part contains the data, for each a(ij) not
    +
    472  // on the diagonal, two computations are done:
    +
    473  // - using a(ij) in its position;
    +
    474  // - using a(ij) in its "transposed" position (applying the ConjTrans to its data)
    +
    475 
    +
    476  {
    +
    477  const LocalTileIndex index_x(Coord::Row, ij.row());
    +
    478  const LocalTileIndex index_w(Coord::Row, ij.col());
    +
    479  hemmOffDiag<B>(thread_priority::high, blas::Op::NoTrans, tile_a, w.read(index_w),
    +
    480  x.readwrite(index_x));
    +
    481  }
    +
    482 
    +
    483  {
    +
    484  const LocalTileIndex index_pretended = transposed(ij);
    +
    485  const LocalTileIndex index_x(Coord::Row, index_pretended.row());
    +
    486  const LocalTileIndex index_w(Coord::Row, index_pretended.col());
    +
    487  hemmOffDiag<B>(thread_priority::high, blas::Op::ConjTrans, tile_a, w.read(index_w),
    +
    488  x.readwrite(index_x));
    +
    489  }
    +
    490  }
    +
    491  }
    +
    492  }
    +
    493 }
    +
    494 
    +
    495 template <Backend B, Device D, class T>
    +
    496 void gemmComputeW2(matrix::Matrix<T, D>& w2, matrix::Panel<Coord::Col, const T, D>& w,
    +
    497  matrix::Panel<Coord::Col, const T, D>& x) {
    +
    498  using pika::execution::thread_priority;
    +
    499 
    +
    500  namespace ex = pika::execution::experimental;
    +
    501 
    +
    502  // Note:
    +
    503  // Not all ranks in the column always hold at least a tile in the panel Ai, but all ranks in
    +
    504  // the column are going to participate to the reduce. For them, it is important to set the
    +
    505  // partial result W2 to zero.
    +
    506  ex::start_detached(w2.readwrite(LocalTileIndex(0, 0)) |
    +
    507  tile::set0(dlaf::internal::Policy<B>(thread_priority::high)));
    +
    508 
    +
    509  using namespace blas;
    +
    510  // GEMM W2 = W* . X
    +
    511  for (const auto& index_tile : w.iteratorLocal())
    +
    512  ex::start_detached(dlaf::internal::whenAllLift(Op::ConjTrans, Op::NoTrans, T(1), w.read(index_tile),
    +
    513  x.read(index_tile), T(1),
    +
    514  w2.readwrite(LocalTileIndex(0, 0))) |
    +
    515  tile::gemm(dlaf::internal::Policy<B>(thread_priority::high)));
    +
    516 }
    +
    517 
    +
    518 template <Backend B, Device D, class T>
    +
    519 void her2kUpdateTrailingMatrix(const matrix::SubMatrixView& view, matrix::Matrix<T, D>& a,
    +
    520  matrix::Panel<Coord::Col, const T, D>& x,
    +
    521  matrix::Panel<Coord::Col, const T, D>& v) {
    +
    522  static_assert(std::is_signed_v<BaseType<T>>, "alpha in computations requires to be -1");
    +
    523 
    +
    524  using pika::execution::thread_priority;
    +
    525 
    +
    526  const auto dist = a.distribution();
    +
    527 
    +
    528  const LocalTileIndex at_start = view.begin();
    +
    529 
    +
    530  for (SizeType i = at_start.row(); i < dist.localNrTiles().rows(); ++i) {
    +
    531  const auto limit = dist.template nextLocalTileFromGlobalTile<Coord::Col>(
    +
    532  dist.template globalTileFromLocalTile<Coord::Row>(i) + 1);
    +
    533  for (SizeType j = at_start.col(); j < limit; ++j) {
    +
    534  const LocalTileIndex ij_local{i, j};
    +
    535  const GlobalTileIndex ij = dist.globalTileIndex(ij_local);
    +
    536 
    +
    537  const bool is_diagonal_tile = (ij.row() == ij.col());
    +
    538 
    +
    539  auto getSubA = [&a, &view, ij_local]() {
    +
    540  return splitTile(a.readwrite(ij_local), view(ij_local));
    +
    541  };
    +
    542 
    +
    543  // The first column of the trailing matrix (except for the very first global tile) has to be
    +
    544  // updated first, in order to unlock the next iteration as soon as possible.
    +
    545  const auto priority = (j == at_start.col()) ? thread_priority::high : thread_priority::normal;
    +
    546 
    +
    547  if (is_diagonal_tile) {
    +
    548  her2kDiag<B>(priority, v.read(ij_local), x.read(ij_local), getSubA());
    +
    549  }
    +
    550  else {
    +
    551  // A -= X . V*
    +
    552  her2kOffDiag<B>(priority, x.read(ij_local), v.read(transposed(ij_local)), getSubA());
    +
    553 
    +
    554  // A -= V . X*
    +
    555  her2kOffDiag<B>(priority, v.read(ij_local), x.read(transposed(ij_local)), getSubA());
    +
    556  }
    +
    557  }
    +
    558  }
    +
    559 }
    +
    560 
    +
    561 }
    +
    562 
    +
    563 namespace distributed {
    +
    564 template <Device D, class T>
    +
    565 T computeReflector(const bool has_head, comm::Communicator& communicator,
    +
    566  const std::vector<matrix::Tile<T, D>>& panel, SizeType j) {
    +
    567  std::array<T, 2> x0_and_squares = computeX0AndSquares(has_head, panel, j);
    +
    568 
    +
    569  // Note:
    +
    570  // This is an optimization for grouping two separate low bandwidth communications, respectively
    +
    571  // bcast(x0) and reduce(norm), where the latency was degrading performances.
    +
    572  //
    +
    573  // In particular this allReduce allows to:
    +
    574  // - bcast x0, since for all ranks is 0 and just the root rank has the real value;
    +
    575  // - allReduce squares for the norm computation.
    +
    576  //
    +
    577  // Moreover, by all-reducing squares and broadcasting x0, all ranks have all the information to
    +
    578  // update locally the reflectors (section they have). This is more efficient than computing params
    +
    579  // (e.g. norm, y, tau) just on the root rank and then having to broadcast them (i.e. additional
    +
    580  // communication).
    +
    581  comm::sync::allReduceInPlace(communicator, MPI_SUM,
    +
    582  common::make_data(x0_and_squares.data(),
    +
    583  to_SizeType(x0_and_squares.size())));
    +
    584 
    +
    585  auto tau = computeReflectorAndTau(has_head, panel, j, std::move(x0_and_squares));
    +
    586 
    +
    587  return tau;
    +
    588 }
    +
    589 
    +
    590 template <class MatrixLikeA, class MatrixLikeTaus, class TriggerSender, class CommSender>
    +
    591 void computePanelReflectors(TriggerSender&& trigger, comm::IndexT_MPI rank_v0,
    +
    592  CommSender&& mpi_col_chain_panel, MatrixLikeA& mat_a,
    +
    593  MatrixLikeTaus& mat_taus, SizeType j_sub,
    +
    594  const matrix::SubPanelView& panel_view) {
    +
    595  static Device constexpr D = MatrixLikeA::device;
    +
    596  using T = typename MatrixLikeA::ElementType;
    +
    597  namespace ex = pika::execution::experimental;
    +
    598  namespace di = dlaf::internal;
    +
    599 
    +
    600  std::vector<matrix::ReadWriteTileSender<T, D>> panel_tiles;
    +
    601  panel_tiles.reserve(to_sizet(std::distance(panel_view.iteratorLocal().begin(),
    +
    602  panel_view.iteratorLocal().end())));
    +
    603  for (const auto& i : panel_view.iteratorLocal()) {
    +
    604  const matrix::SubTileSpec& spec = panel_view(i);
    +
    605  panel_tiles.emplace_back(matrix::splitTile(mat_a.readwrite(i), spec));
    +
    606  }
    +
    607 
    +
    608  const std::size_t nthreads = getReductionToBandPanelNWorkers();
    +
    609  auto s =
    +
    610  ex::when_all(ex::just(std::make_unique<pika::barrier<>>(nthreads),
    +
    611  std::vector<common::internal::vector<T>>{}), // w (internally required)
    +
    612  mat_taus.readwrite(GlobalTileIndex(j_sub, 0)),
    +
    613  ex::when_all_vector(std::move(panel_tiles)),
    +
    614  std::forward<CommSender>(mpi_col_chain_panel), std::forward<TriggerSender>(trigger)) |
    +
    615  ex::transfer(di::getBackendScheduler<Backend::MC>(pika::execution::thread_priority::high)) |
    +
    616  ex::bulk(nthreads, [nthreads, rank_v0,
    +
    617  cols = panel_view.cols()](const std::size_t index, auto& barrier_ptr, auto& w,
    +
    618  auto& taus, auto& tiles, auto&& pcomm) {
    +
    619  const bool rankHasHead = rank_v0 == pcomm.get().rank();
    +
    620 
    +
    621  const auto barrier_busy_wait = getReductionToBandBarrierBusyWait();
    +
    622  const std::size_t batch_size = util::ceilDiv(tiles.size(), nthreads);
    +
    623  const std::size_t begin = index * batch_size;
    +
    624  const std::size_t end = std::min(index * batch_size + batch_size, tiles.size());
    +
    625  const SizeType nrefls = taus.size().rows();
    +
    626 
    +
    627  if (index == 0) {
    +
    628  w.resize(nthreads);
    +
    629  }
    +
    630 
    +
    631  for (SizeType j = 0; j < nrefls; ++j) {
    +
    632  // STEP1: compute tau and reflector (single-thread)
    +
    633  if (index == 0) {
    +
    634  const bool has_head = rankHasHead;
    +
    635  taus({j, 0}) = computeReflector(has_head, pcomm.get(), tiles, j);
    +
    636  }
    +
    637  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    +
    638 
    +
    639  // STEP2a: compute w (multi-threaded)
    +
    640  const SizeType pt_cols = cols - (j + 1);
    +
    641  if (pt_cols == 0)
    +
    642  break;
    +
    643 
    +
    644  const bool has_head = rankHasHead && (index == 0);
    +
    645 
    +
    646  w[index] = common::internal::vector<T>(pt_cols, 0);
    +
    647  computeWTrailingPanel(has_head, tiles, w[index], j, pt_cols, begin, end);
    +
    648  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    +
    649 
    +
    650  // STEP2b: reduce w results (single-threaded)
    +
    651  if (index == 0) {
    +
    652  dlaf::eigensolver::internal::reduceColumnVectors(w);
    +
    653  comm::sync::allReduceInPlace(pcomm.get(), MPI_SUM, common::make_data(w[0].data(), pt_cols));
    +
    654  }
    +
    655  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    +
    656 
    +
    657  // STEP3: update trailing panel (multi-threaded)
    +
    658  updateTrailingPanel(has_head, tiles, j, w[0], taus({j, 0}), begin, end);
    +
    659  barrier_ptr->arrive_and_wait(barrier_busy_wait);
    +
    660  }
    +
    661  });
    +
    662  ex::start_detached(std::move(s));
    +
    663 }
    +
    664 
    +
    665 template <Backend B, Device D, class T>
    +
    666 void hemmComputeX(comm::IndexT_MPI reducer_col, matrix::Panel<Coord::Col, T, D>& x,
    +
    667  matrix::Panel<Coord::Row, T, D, matrix::StoreTransposed::Yes>& xt,
    +
    668  const matrix::SubMatrixView& view, matrix::Matrix<const T, D>& a,
    +
    669  matrix::Panel<Coord::Col, const T, D>& w,
    +
    670  matrix::Panel<Coord::Row, const T, D, matrix::StoreTransposed::Yes>& wt,
    +
    671  common::Pipeline<comm::Communicator>& mpi_row_chain,
    +
    672  common::Pipeline<comm::Communicator>& mpi_col_chain) {
    +
    673  namespace ex = pika::execution::experimental;
    +
    674 
    +
    675  using pika::execution::thread_priority;
    +
    676 
    +
    677  const auto dist = a.distribution();
    +
    678  const auto rank = dist.rankIndex();
    +
    679 
    +
    680  // Note:
    +
    681  // They have to be set to zero, because all tiles are going to be reduced, and some tiles may not get
    +
    682  // "initialized" during computation, so they should not contribute with any spurious value to the final
    +
    683  // result.
    +
    684  matrix::util::set0<B>(thread_priority::high, x);
    +
    685  matrix::util::set0<B>(thread_priority::high, xt);
    +
    686 
    +
    687  const LocalTileIndex at_offset = view.begin();
    +
    688 
    +
    689  for (SizeType i = at_offset.row(); i < dist.localNrTiles().rows(); ++i) {
    +
    690  const auto limit = dist.template nextLocalTileFromGlobalTile<Coord::Col>(
    +
    691  dist.template globalTileFromLocalTile<Coord::Row>(i) + 1);
    +
    692  for (SizeType j = limit - 1; j >= at_offset.col(); --j) {
    +
    693  const LocalTileIndex ij_local{i, j};
    +
    694  const GlobalTileIndex ij = dist.globalTileIndex(ij_local);
    +
    695 
    +
    696  const bool is_diagonal_tile = (ij.row() == ij.col());
    +
    697 
    +
    698  auto tile_a = splitTile(a.read(ij), view(ij_local));
    +
    699 
    +
    700  if (is_diagonal_tile) {
    +
    701  hemmDiag<B>(thread_priority::high, std::move(tile_a), w.read(ij_local), x.readwrite(ij_local));
    +
    702  }
    +
    703  else {
    +
    704  // Note:
    +
    705  // Since it is not a diagonal tile, otherwise it would have been managed in the previous
    +
    706  // branch, the second operand is not available in W but it is accessible through the
    +
    707  // support panel Wt.
    +
    708  // However, since we are still computing the "straight" part, the result can be stored
    +
    709  // in the "local" panel X.
    +
    710  hemmOffDiag<B>(thread_priority::high, blas::Op::NoTrans, tile_a, wt.read(ij_local),
    +
    711  x.readwrite(ij_local));
    +
    712 
    +
    713  // Note:
    +
    714  // Here we are considering the hermitian part of A, so coordinates have to be "mirrored".
    +
    715  // So, first step is identifying the mirrored cell coordinate, i.e. swap row/col, together
    +
    716  // with realizing if the new coord lays on an owned row or not.
    +
    717  // If yes, the result can be stored in the X, otherwise Xt support panel will be used.
    +
    718  // For what concerns the second operand, it can be found for sure in W. In fact, the
    +
    719  // multiplication requires matching col(A) == row(W), but since coordinates are mirrored,
    +
    720  // we are matching row(A) == row(W), so it is local by construction.
    +
    721  const auto owner = dist.template rankGlobalTile<Coord::Row>(ij.col());
    +
    722 
    +
    723  const LocalTileIndex index_x{dist.template localTileFromGlobalTile<Coord::Row>(ij.col()), 0};
    +
    724  const LocalTileIndex index_xt{0, ij_local.col()};
    +
    725 
    +
    726  auto tile_x = (dist.rankIndex().row() == owner) ? x.readwrite(index_x) : xt.readwrite(index_xt);
    +
    727 
    +
    728  hemmOffDiag<B>(thread_priority::high, blas::Op::ConjTrans, std::move(tile_a), w.read(ij_local),
    +
    729  std::move(tile_x));
    +
    730  }
    +
    731  }
    +
    732  }
    +
    733 
    +
    734  // Note:
    +
    735  // At this point, partial results of X and Xt are available in the panels, and they have to be reduced,
    +
    736  // both row-wise and col-wise.
    +
    737  // The final X result will be available just on Ai panel column.
    +
    738 
    +
    739  // Note:
    +
    740  // The first step in reducing partial results distributed over X and Xt, it is to reduce the row
    +
    741  // panel Xt col-wise, by collecting all Xt results on the rank which can "mirror" the result on its
    +
    742  // rows (i.e. diagonal). So, for each tile of the row panel, select who is the "diagonal" rank that can
    +
    743  // mirror and reduce on it.
    +
    744  for (const auto& index_xt : xt.iteratorLocal()) {
    +
    745  const auto index_k = dist.template globalTileFromLocalTile<Coord::Col>(index_xt.col());
    +
    746  const auto rank_owner_row = dist.template rankGlobalTile<Coord::Row>(index_k);
    +
    747 
    +
    748  if (rank_owner_row == rank.row()) {
    +
    749  // Note:
    +
    750  // Since it is the owner, it has to perform the "mirroring" of the results from columns to
    +
    751  // rows.
    +
    752  //
    +
    753  // Moreover, it reduces in place because the owner of the diagonal stores the partial result
    +
    754  // directly in x (without using xt)
    +
    755  const auto i = dist.template localTileFromGlobalTile<Coord::Row>(index_k);
    +
    756  ex::start_detached(comm::scheduleReduceRecvInPlace(mpi_col_chain(), MPI_SUM, x.readwrite({i, 0})));
    +
    757  }
    +
    758  else {
    +
    759  ex::start_detached(comm::scheduleReduceSend(mpi_col_chain(), rank_owner_row, MPI_SUM,
    +
    760  xt.read(index_xt)));
    +
    761  }
    +
    762  }
    +
    763 
    +
    764  // Note:
    +
    765  // At this point partial results are all collected in X (Xt has been embedded in previous step),
    +
    766  // so the last step needed is to reduce these last partial results in the final results.
    +
    767  // The result is needed just on the column with reflectors.
    +
    768  for (const auto& index_x : x.iteratorLocal()) {
    +
    769  if (reducer_col == rank.col())
    +
    770  ex::start_detached(comm::scheduleReduceRecvInPlace(mpi_row_chain(), MPI_SUM,
    +
    771  x.readwrite(index_x)));
    +
    772  else
    +
    773  ex::start_detached(comm::scheduleReduceSend(mpi_row_chain(), reducer_col, MPI_SUM,
    +
    774  x.read(index_x)));
    +
    775  }
    +
    776 }
    +
    777 
    +
    778 template <Backend B, Device D, class T>
    +
    779 void her2kUpdateTrailingMatrix(const matrix::SubMatrixView& view, Matrix<T, D>& a,
    +
    780  matrix::Panel<Coord::Col, const T, D>& x,
    +
    781  matrix::Panel<Coord::Row, const T, D, matrix::StoreTransposed::Yes>& vt,
    +
    782  matrix::Panel<Coord::Col, const T, D>& v,
    +
    783  matrix::Panel<Coord::Row, const T, D, matrix::StoreTransposed::Yes>& xt) {
    +
    784  static_assert(std::is_signed_v<BaseType<T>>, "alpha in computations requires to be -1");
    +
    785 
    +
    786  using pika::execution::thread_priority;
    +
    787 
    +
    788  const auto dist = a.distribution();
    +
    789 
    +
    790  const LocalTileIndex at_start = view.begin();
    +
    791 
    +
    792  for (SizeType i = at_start.row(); i < dist.localNrTiles().rows(); ++i) {
    +
    793  const auto limit = dist.template nextLocalTileFromGlobalTile<Coord::Col>(
    +
    794  dist.template globalTileFromLocalTile<Coord::Row>(i) + 1);
    +
    795  for (SizeType j = at_start.col(); j < limit; ++j) {
    +
    796  const LocalTileIndex ij_local{i, j};
    +
    797  const GlobalTileIndex ij = dist.globalTileIndex(ij_local);
    +
    798 
    +
    799  const bool is_diagonal_tile = (ij.row() == ij.col());
    +
    800 
    +
    801  auto getSubA = [&a, &view, ij_local]() {
    +
    802  return splitTile(a.readwrite(ij_local), view(ij_local));
    +
    803  };
    +
    804 
    +
    805  // The first column of the trailing matrix (except for the very first global tile) has to be
    +
    806  // updated first, in order to unlock the next iteration as soon as possible.
    +
    807  const auto priority = (j == at_start.col()) ? thread_priority::high : thread_priority::normal;
    +
    808 
    +
    809  if (is_diagonal_tile) {
    +
    810  her2kDiag<B>(priority, v.read(ij_local), x.read(ij_local), getSubA());
    +
    811  }
    +
    812  else {
    +
    813  // A -= X . V*
    +
    814  her2kOffDiag<B>(priority, x.read(ij_local), vt.read(ij_local), getSubA());
    +
    815 
    +
    816  // A -= V . X*
    +
    817  her2kOffDiag<B>(priority, v.read(ij_local), xt.read(ij_local), getSubA());
    +
    818  }
    +
    819  }
    +
    820  }
    +
    821 }
    822 }
    -
    823 }
    -
    824 
    -
    825 template <Backend B, Device D, class T>
    -
    826 struct ComputePanelHelper;
    -
    827 
    -
    828 template <class T>
    -
    829 struct ComputePanelHelper<Backend::MC, Device::CPU, T> {
    -
    830  ComputePanelHelper(const std::size_t, matrix::Distribution) {}
    -
    831 
    -
    832  void call(Matrix<T, Device::CPU>& mat_a, matrix::RetiledMatrix<T, Device::CPU>& mat_taus,
    -
    833  const SizeType j_sub, const matrix::SubPanelView& panel_view) {
    -
    834  using red2band::local::computePanelReflectors;
    -
    835  computePanelReflectors(mat_a, mat_taus, j_sub, panel_view);
    -
    836  }
    -
    837 
    -
    838  template <Device D, class CommSender, class TriggerSender>
    -
    839  void call(TriggerSender&& trigger, comm::IndexT_MPI rank_v0, CommSender&& mpi_col_chain_panel,
    -
    840  Matrix<T, D>& mat_a, matrix::RetiledMatrix<T, Device::CPU>& mat_taus, const SizeType j_sub,
    -
    841  const matrix::SubPanelView& panel_view) {
    -
    842  using red2band::distributed::computePanelReflectors;
    -
    843  computePanelReflectors(std::forward<TriggerSender>(trigger), rank_v0,
    -
    844  std::forward<CommSender>(mpi_col_chain_panel), mat_a, mat_taus, j_sub,
    -
    845  panel_view);
    -
    846  }
    -
    847 };
    -
    848 
    -
    849 #ifdef DLAF_WITH_GPU
    -
    850 template <class T>
    -
    851 struct ComputePanelHelper<Backend::GPU, Device::GPU, T> {
    -
    852  ComputePanelHelper(const std::size_t n_workspaces, matrix::Distribution dist_a)
    -
    853  : panels_v(n_workspaces, dist_a) {}
    -
    854 
    -
    855  void call(Matrix<T, Device::GPU>& mat_a, matrix::RetiledMatrix<T, Device::CPU>& mat_taus,
    -
    856  const SizeType j_sub, const matrix::SubPanelView& panel_view) {
    -
    857  using red2band::local::computePanelReflectors;
    -
    858 
    -
    859  namespace ex = pika::execution::experimental;
    -
    860 
    -
    861  // Note:
    -
    862  // - copy panel_view from GPU to CPU
    -
    863  // - computePanelReflectors on CPU (on a matrix like, with just a panel)
    -
    864  // - copy back matrix "panel" from CPU to GPU
    -
    865 
    -
    866  auto& v = panels_v.nextResource();
    -
    867 
    -
    868  copyToCPU(panel_view, mat_a, v);
    -
    869  computePanelReflectors(v, mat_taus, j_sub, panel_view);
    -
    870  copyFromCPU(panel_view, v, mat_a);
    -
    871  }
    -
    872 
    -
    873  template <Device D, class CommSender, class TriggerSender>
    -
    874  void call(TriggerSender&& trigger, comm::IndexT_MPI rank_v0, CommSender&& mpi_col_chain_panel,
    -
    875  Matrix<T, D>& mat_a, matrix::RetiledMatrix<T, Device::CPU>& mat_taus, SizeType j_sub,
    -
    876  const matrix::SubPanelView& panel_view) {
    -
    877  auto& v = panels_v.nextResource();
    -
    878 
    -
    879  // copy to CPU
    -
    880  copyToCPU(panel_view, mat_a, v);
    -
    881 
    -
    882  // compute on CPU
    -
    883  using dlaf::eigensolver::internal::red2band::distributed::computePanelReflectors;
    -
    884  computePanelReflectors(std::forward<TriggerSender>(trigger), rank_v0,
    -
    885  std::forward<CommSender>(mpi_col_chain_panel), v, mat_taus, j_sub,
    -
    886  panel_view);
    -
    887 
    -
    888  // copy back to GPU
    -
    889  copyFromCPU(panel_view, v, mat_a);
    -
    890  }
    -
    891 
    -
    892 protected:
    -
    893  common::RoundRobin<matrix::Panel<Coord::Col, T, Device::CPU>> panels_v;
    -
    894 
    -
    895  void copyToCPU(const matrix::SubPanelView panel_view, matrix::Matrix<T, Device::GPU>& mat_a,
    -
    896  matrix::Panel<Coord::Col, T, Device::CPU>& v) {
    -
    897  namespace ex = pika::execution::experimental;
    -
    898 
    -
    899  using dlaf::internal::Policy;
    -
    900  using dlaf::matrix::internal::CopyBackend_v;
    -
    901  using pika::execution::thread_priority;
    -
    902 
    -
    903  for (const auto& i : panel_view.iteratorLocal()) {
    -
    904  auto spec = panel_view(i);
    -
    905  auto tmp_tile = v.readwrite(i);
    -
    906  ex::start_detached(
    -
    907  ex::when_all(splitTile(mat_a.read(i), spec), splitTile(std::move(tmp_tile), spec)) |
    -
    908  matrix::copy(Policy<CopyBackend_v<Device::GPU, Device::CPU>>(thread_priority::high)));
    -
    909  }
    -
    910  }
    -
    911 
    -
    912  void copyFromCPU(const matrix::SubPanelView panel_view, matrix::Panel<Coord::Col, T, Device::CPU>& v,
    -
    913  matrix::Matrix<T, Device::GPU>& mat_a) {
    -
    914  namespace ex = pika::execution::experimental;
    -
    915 
    -
    916  using dlaf::internal::Policy;
    -
    917  using dlaf::matrix::internal::CopyBackend_v;
    -
    918  using pika::execution::thread_priority;
    -
    919 
    -
    920  for (const auto& i : panel_view.iteratorLocal()) {
    -
    921  auto spec = panel_view(i);
    -
    922  auto tile_a = mat_a.readwrite(i);
    -
    923  ex::start_detached(
    -
    924  ex::when_all(splitTile(v.read(i), spec), splitTile(std::move(tile_a), spec)) |
    -
    925  matrix::copy(Policy<CopyBackend_v<Device::CPU, Device::GPU>>(thread_priority::high)));
    -
    926  }
    -
    927  }
    -
    928 };
    -
    929 #endif
    -
    930 
    -
    931 }
    -
    932 
    -
    933 // Local implementation of reduction to band
    -
    934 template <Backend B, Device D, class T>
    -
    935 Matrix<T, Device::CPU> ReductionToBand<B, D, T>::call(Matrix<T, D>& mat_a, const SizeType band_size) {
    -
    936  using dlaf::matrix::Matrix;
    -
    937  using dlaf::matrix::Panel;
    -
    938 
    -
    939  using namespace red2band::local;
    -
    940 
    -
    941  using common::iterate_range2d;
    -
    942  using factorization::internal::computeTFactor;
    -
    943 
    -
    944  using pika::execution::experimental::any_sender;
    -
    945 
    -
    946  const auto dist_a = mat_a.distribution();
    -
    947  const matrix::Distribution dist({mat_a.size().rows(), band_size},
    -
    948  {dist_a.blockSize().rows(), band_size});
    -
    949 
    -
    950  // Note:
    -
    951  // Reflector of size = 1 is not considered whatever T is (i.e. neither real nor complex)
    -
    952  const SizeType nrefls = std::max<SizeType>(0, dist_a.size().rows() - band_size - 1);
    -
    953 
    -
    954  // Row-vector that is distributed over columns, but exists locally on all rows of the grid
    -
    955  DLAF_ASSERT(mat_a.blockSize().cols() % band_size == 0, mat_a.blockSize().cols(), band_size);
    -
    956  Matrix<T, Device::CPU> mat_taus(matrix::Distribution(GlobalElementSize(nrefls, 1),
    -
    957  TileElementSize(mat_a.blockSize().cols(), 1),
    -
    958  comm::Size2D(mat_a.commGridSize().cols(), 1),
    -
    959  comm::Index2D(mat_a.rankIndex().col(), 0),
    -
    960  comm::Index2D(mat_a.sourceRankIndex().col(), 0)));
    -
    961 
    -
    962  if (nrefls == 0)
    -
    963  return mat_taus;
    -
    964 
    -
    965  matrix::RetiledMatrix<T, Device::CPU> mat_taus_retiled(
    -
    966  mat_taus, LocalTileSize(mat_a.blockSize().cols() / band_size, 1));
    -
    967 
    -
    968  const SizeType ntiles = (nrefls - 1) / band_size + 1;
    -
    969  DLAF_ASSERT(ntiles == mat_taus_retiled.nrTiles().rows(), ntiles, mat_taus_retiled.nrTiles().rows());
    -
    970 
    -
    971  const bool is_full_band = (band_size == dist_a.blockSize().cols());
    -
    972 
    -
    973  constexpr std::size_t n_workspaces = 2;
    -
    974  common::RoundRobin<Panel<Coord::Col, T, D>> panels_v(n_workspaces, dist);
    -
    975  common::RoundRobin<Panel<Coord::Col, T, D>> panels_w(n_workspaces, dist);
    -
    976  common::RoundRobin<Panel<Coord::Col, T, D>> panels_x(n_workspaces, dist);
    -
    977 
    -
    978  // Note:
    -
    979  // Here dist_a is given with full panel size instead of dist with just the part actually needeed,
    -
    980  // because the GPU Helper internally exploits Panel data-structure. Indeed, the full size panel is
    -
    981  // needed in order to mimick Matrix with Panel, so it is possible to apply a SubPanelView to it.
    -
    982  //
    -
    983  // It is a bit hacky usage, because SubPanelView is not meant to be used with Panel, but just with
    -
    984  // Matrix. This results in a variable waste of memory, depending no the ratio band_size/nb.
    -
    985  red2band::ComputePanelHelper<B, D, T> compute_panel_helper(n_workspaces, dist_a);
    -
    986 
    -
    987  for (SizeType j_sub = 0; j_sub < ntiles; ++j_sub) {
    -
    988  const auto i_sub = j_sub + 1;
    -
    989 
    -
    990  const GlobalElementIndex ij_offset(i_sub * band_size, j_sub * band_size);
    -
    991 
    -
    992  const SizeType nrefls_tile = mat_taus_retiled.tileSize(GlobalTileIndex(j_sub, 0)).rows();
    -
    993 
    -
    994  const bool isPanelIncomplete = (nrefls_tile != band_size);
    -
    995 
    -
    996  // Note: if this is running, it must have at least one valid reflector (i.e. with size > 1)
    -
    997  DLAF_ASSERT_HEAVY(nrefls_tile != 0, nrefls_tile);
    -
    998 
    -
    999  // Note: SubPanelView is (at most) band_size wide, but it may contain a smaller number of
    -
    1000  // reflectors (i.e. at the end when last reflector size is 1)
    -
    1001  const matrix::SubPanelView panel_view(dist_a, ij_offset, band_size);
    -
    1002 
    -
    1003  Panel<Coord::Col, T, D>& v = panels_v.nextResource();
    -
    1004  v.setRangeStart(ij_offset);
    -
    1005  if (isPanelIncomplete)
    -
    1006  v.setWidth(nrefls_tile);
    -
    1007 
    -
    1008  // PANEL
    -
    1009  compute_panel_helper.call(mat_a, mat_taus_retiled, j_sub, panel_view);
    -
    1010 
    -
    1011  // Note:
    -
    1012  // - has_reflector_head tells if this rank owns the first tile of the panel (being local, always true)
    -
    1013  // - if !is_full_band it has to force copy as a workaround, otherwise in update matrix it would deadlock
    -
    1014  // due to tile shared between panel and trailing matrix
    -
    1015  constexpr bool has_reflector_head = true;
    -
    1016  setupReflectorPanelV<B, D, T>(has_reflector_head, panel_view, nrefls_tile, v, mat_a, !is_full_band);
    -
    1017 
    -
    1018  const LocalTileIndex t_idx(0, 0);
    -
    1019  // TODO used just by the column, maybe we can re-use a panel tile?
    -
    1020  // TODO probably the first one in any panel is ok?
    -
    1021  Matrix<T, D> t({nrefls_tile, nrefls_tile}, dist.blockSize());
    -
    1022 
    -
    1023  computeTFactor<B>(v, mat_taus_retiled.read(GlobalTileIndex(j_sub, 0)), t.readwrite(t_idx));
    -
    1024 
    -
    1025  // PREPARATION FOR TRAILING MATRIX UPDATE
    -
    1026  const GlobalElementIndex at_offset(ij_offset + GlobalElementSize(0, band_size));
    -
    1027 
    -
    1028  // Note: if there is no trailing matrix, algorithm has finised
    -
    1029  if (!at_offset.isIn(mat_a.size()))
    -
    1030  break;
    -
    1031 
    -
    1032  const matrix::SubMatrixView trailing_matrix_view(dist_a, at_offset);
    -
    1033 
    -
    1034  // W = V . T
    -
    1035  Panel<Coord::Col, T, D>& w = panels_w.nextResource();
    -
    1036  w.setRangeStart(at_offset);
    -
    1037  if (isPanelIncomplete)
    -
    1038  w.setWidth(nrefls_tile);
    -
    1039 
    -
    1040  trmmComputeW<B>(w, v, t.read(t_idx));
    -
    1041 
    -
    1042  // X = At . W
    -
    1043  Panel<Coord::Col, T, D>& x = panels_x.nextResource();
    -
    1044  x.setRangeStart(at_offset);
    -
    1045  if (isPanelIncomplete)
    -
    1046  x.setWidth(nrefls_tile);
    -
    1047 
    -
    1048  // Note:
    -
    1049  // Since At is hermitian, just the lower part is referenced.
    -
    1050  // When the tile is not part of the main diagonal, the same tile has to be used for two computations
    -
    1051  // that will contribute to two different rows of X: the ones indexed with row and col.
    -
    1052  hemmComputeX<B>(x, trailing_matrix_view, mat_a, w);
    -
    1053 
    -
    1054  // In the next section the next two operations are performed
    -
    1055  // A) W2 = W* . X
    -
    1056  // B) X -= 1/2 . V . W2
    -
    1057 
    -
    1058  // Note:
    -
    1059  // T can be re-used because it is not needed anymore in this step and it has the same shape
    -
    1060  Matrix<T, D> w2 = std::move(t);
    -
    1061 
    -
    1062  gemmComputeW2<B>(w2, w, x);
    -
    1063  gemmUpdateX<B>(x, w2, v);
    -
    1064 
    -
    1065  // TRAILING MATRIX UPDATE
    -
    1066 
    -
    1067  // At -= X . V* + V . X*
    -
    1068  her2kUpdateTrailingMatrix<B>(trailing_matrix_view, mat_a, x, v);
    -
    1069 
    -
    1070  x.reset();
    -
    1071  w.reset();
    -
    1072  v.reset();
    -
    1073  }
    -
    1074 
    -
    1075  return mat_taus;
    -
    1076 }
    -
    1077 
    -
    1078 // Distributed implementation of reduction to band
    -
    1079 template <Backend B, Device D, class T>
    -
    1080 Matrix<T, Device::CPU> ReductionToBand<B, D, T>::call(comm::CommunicatorGrid grid, Matrix<T, D>& mat_a,
    -
    1081  const SizeType band_size) {
    -
    1082  using namespace red2band::distributed;
    -
    1083 
    -
    1084  using common::iterate_range2d;
    -
    1085  using factorization::internal::computeTFactor;
    -
    1086 
    -
    1087  namespace ex = pika::execution::experimental;
    -
    1088 
    -
    1089  // Note:
    -
    1090  // This is a temporary workaround.
    -
    1091  // See issue https://github.com/eth-cscs/DLA-Future/issues/729
    -
    1092  pika::threads::get_thread_manager().wait();
    -
    1093 
    -
    1094  common::Pipeline<comm::Communicator> mpi_col_chain_panel(grid.colCommunicator().clone());
    -
    1095  common::Pipeline<comm::Communicator> mpi_row_chain(grid.rowCommunicator().clone());
    -
    1096  common::Pipeline<comm::Communicator> mpi_col_chain(grid.colCommunicator().clone());
    -
    1097 
    -
    1098  const auto& dist = mat_a.distribution();
    -
    1099  const comm::Index2D rank = dist.rankIndex();
    -
    1100 
    -
    1101  // Note:
    -
    1102  // Reflector of size = 1 is not considered whatever T is (i.e. neither real nor complex)
    -
    1103  const SizeType nrefls = std::max<SizeType>(0, dist.size().rows() - band_size - 1);
    -
    1104 
    -
    1105  // Row-vector that is distributed over columns, but exists locally on all rows of the grid
    -
    1106  DLAF_ASSERT(mat_a.blockSize().cols() % band_size == 0, mat_a.blockSize().cols(), band_size);
    -
    1107  Matrix<T, Device::CPU> mat_taus(matrix::Distribution(GlobalElementSize(nrefls, 1),
    -
    1108  TileElementSize(mat_a.blockSize().cols(), 1),
    -
    1109  comm::Size2D(mat_a.commGridSize().cols(), 1),
    -
    1110  comm::Index2D(mat_a.rankIndex().col(), 0),
    -
    1111  comm::Index2D(mat_a.sourceRankIndex().col(), 0)));
    -
    1112 
    -
    1113  if (nrefls == 0)
    -
    1114  return mat_taus;
    -
    1115 
    -
    1116  matrix::RetiledMatrix<T, Device::CPU> mat_taus_retiled(
    -
    1117  mat_taus, LocalTileSize(mat_a.blockSize().cols() / band_size, 1));
    -
    1118 
    -
    1119  const SizeType ntiles = (nrefls - 1) / band_size + 1;
    -
    1120  DLAF_ASSERT(ntiles == mat_taus_retiled.nrTiles().rows(), ntiles, mat_taus_retiled.nrTiles().rows());
    -
    1121 
    -
    1122  const bool is_full_band = (band_size == dist.blockSize().cols());
    -
    1123 
    -
    1124  constexpr std::size_t n_workspaces = 2;
    -
    1125  common::RoundRobin<matrix::Panel<Coord::Col, T, D>> panels_v(n_workspaces, dist);
    -
    1126  common::RoundRobin<matrix::Panel<Coord::Row, T, D, matrix::StoreTransposed::Yes>> panels_vt(
    -
    1127  n_workspaces, dist);
    -
    1128 
    -
    1129  common::RoundRobin<matrix::Panel<Coord::Col, T, D>> panels_w(n_workspaces, dist);
    -
    1130  common::RoundRobin<matrix::Panel<Coord::Row, T, D, matrix::StoreTransposed::Yes>> panels_wt(
    -
    1131  n_workspaces, dist);
    -
    1132 
    -
    1133  common::RoundRobin<matrix::Panel<Coord::Col, T, D>> panels_x(n_workspaces, dist);
    -
    1134  common::RoundRobin<matrix::Panel<Coord::Row, T, D, matrix::StoreTransposed::Yes>> panels_xt(
    -
    1135  n_workspaces, dist);
    -
    1136 
    -
    1137  red2band::ComputePanelHelper<B, D, T> compute_panel_helper(n_workspaces, dist);
    -
    1138 
    -
    1139  ex::unique_any_sender<> trigger_panel{ex::just()};
    -
    1140  for (SizeType j_sub = 0; j_sub < ntiles; ++j_sub) {
    -
    1141  const SizeType i_sub = j_sub + 1;
    -
    1142 
    -
    1143  const GlobalElementIndex ij_offset(i_sub * band_size, j_sub * band_size);
    -
    1144  const GlobalElementIndex at_offset(i_sub * band_size, (j_sub + 1) * band_size);
    -
    1145 
    -
    1146  const comm::Index2D rank_v0{
    -
    1147  dist.template rankGlobalElement<Coord::Row>(ij_offset.row()),
    -
    1148  dist.template rankGlobalElement<Coord::Col>(ij_offset.col()),
    -
    1149  };
    -
    1150 
    -
    1151  const bool is_panel_rank_col = rank_v0.col() == rank.col();
    -
    1152 
    -
    1153  const SizeType nrefls_tile = mat_taus_retiled.tileSize(GlobalTileIndex(j_sub, 0)).rows();
    -
    1154 
    -
    1155  if (nrefls_tile == 0)
    -
    1156  break;
    -
    1157 
    -
    1158  auto& v = panels_v.nextResource();
    -
    1159  auto& vt = panels_vt.nextResource();
    -
    1160 
    -
    1161  v.setRangeStart(at_offset);
    -
    1162  vt.setRangeStart(at_offset);
    -
    1163 
    -
    1164  v.setWidth(nrefls_tile);
    -
    1165  vt.setHeight(nrefls_tile);
    -
    1166 
    -
    1167  const LocalTileIndex t_idx(0, 0);
    -
    1168  // TODO used just by the column, maybe we can re-use a panel tile?
    -
    1169  // TODO or we can keep just the sh_future and allocate just inside if (is_panel_rank_col)
    -
    1170  matrix::Matrix<T, D> t({nrefls_tile, nrefls_tile}, dist.blockSize());
    -
    1171 
    -
    1172  // PANEL
    -
    1173  const matrix::SubPanelView panel_view(dist, ij_offset, band_size);
    -
    1174 
    -
    1175  if (is_panel_rank_col) {
    -
    1176  compute_panel_helper.call(std::move(trigger_panel), rank_v0.row(), mpi_col_chain_panel(), mat_a,
    -
    1177  mat_taus_retiled, j_sub, panel_view);
    -
    1178 
    -
    1179  // Note:
    -
    1180  // - has_reflector_head tells if this rank owns the first tile of the panel
    -
    1181  // - if !is_full_band it has to force copy as a workaround, otherwise in update matrix it would
    -
    1182  // deadlock due to tile shared between panel and trailing matrix
    -
    1183  red2band::local::setupReflectorPanelV<B, D, T>(rank.row() == rank_v0.row(), panel_view,
    -
    1184  nrefls_tile, v, mat_a, !is_full_band);
    -
    1185  computeTFactor<B>(v, mat_taus_retiled.read(GlobalTileIndex(j_sub, 0)), t.readwrite(t_idx),
    -
    1186  mpi_col_chain);
    -
    1187  }
    -
    1188 
    -
    1189  // PREPARATION FOR TRAILING MATRIX UPDATE
    -
    1190 
    -
    1191  // Note: if there is no trailing matrix, algorithm has finised
    -
    1192  if (!at_offset.isIn(mat_a.size()))
    -
    1193  break;
    -
    1194 
    -
    1195  const matrix::SubMatrixView trailing_matrix_view(dist, at_offset);
    -
    1196 
    -
    1197  comm::broadcast(rank_v0.col(), v, vt, mpi_row_chain, mpi_col_chain);
    -
    1198 
    -
    1199  // W = V . T
    -
    1200  auto& w = panels_w.nextResource();
    -
    1201  auto& wt = panels_wt.nextResource();
    -
    1202 
    -
    1203  w.setRangeStart(at_offset);
    -
    1204  wt.setRangeStart(at_offset);
    -
    1205 
    -
    1206  w.setWidth(nrefls_tile);
    -
    1207  wt.setHeight(nrefls_tile);
    -
    1208 
    -
    1209  if (is_panel_rank_col)
    -
    1210  red2band::local::trmmComputeW<B, D>(w, v, t.read(t_idx));
    -
    1211 
    -
    1212  comm::broadcast(rank_v0.col(), w, wt, mpi_row_chain, mpi_col_chain);
    -
    1213 
    -
    1214  // X = At . W
    -
    1215  auto& x = panels_x.nextResource();
    -
    1216  auto& xt = panels_xt.nextResource();
    -
    1217 
    -
    1218  x.setRangeStart(at_offset);
    -
    1219  xt.setRangeStart(at_offset);
    -
    1220 
    -
    1221  x.setWidth(nrefls_tile);
    -
    1222  xt.setHeight(nrefls_tile);
    -
    1223 
    -
    1224  // Note:
    -
    1225  // Since At is hermitian, just the lower part is referenced.
    -
    1226  // When the tile is not part of the main diagonal, the same tile has to be used for two computations
    -
    1227  // that will contribute to two different rows of X: the ones indexed with row and col.
    -
    1228  // This is achieved by storing the two results in two different workspaces: X and X_conj respectively.
    -
    1229  //
    -
    1230  // On exit, x will contain a valid result just on ranks belonging to the column panel.
    -
    1231  // For what concerns xt, it is just used as support and it contains junk data on all ranks.
    -
    1232  hemmComputeX<B, D>(rank_v0.col(), x, xt, trailing_matrix_view, mat_a, w, wt, mpi_row_chain,
    -
    1233  mpi_col_chain);
    -
    1234 
    -
    1235  // In the next section the next two operations are performed
    -
    1236  // A) W2 = W* . X
    -
    1237  // B) X -= 1/2 . V . W2
    -
    1238 
    -
    1239  // Note:
    -
    1240  // Now the intermediate result for X is available on the panel column ranks,
    -
    1241  // which have locally all the needed stuff for updating X and finalize the result
    -
    1242  if (is_panel_rank_col) {
    -
    1243  // Note:
    -
    1244  // T can be re-used because it is not needed anymore in this step and it has the same shape
    -
    1245  matrix::Matrix<T, D> w2 = std::move(t);
    -
    1246 
    -
    1247  red2band::local::gemmComputeW2<B, D>(w2, w, x);
    -
    1248  ex::start_detached(comm::scheduleAllReduceInPlace(mpi_col_chain(), MPI_SUM,
    -
    1249  w2.readwrite(LocalTileIndex(0, 0))));
    -
    1250 
    -
    1251  red2band::local::gemmUpdateX<B, D>(x, w2, v);
    -
    1252  }
    -
    1253 
    -
    1254  // Note:
    -
    1255  // xt has been used previously as workspace for hemmComputeX, so it has to be reset, because now it
    -
    1256  // will be used for accessing the broadcasted version of x
    -
    1257  xt.reset();
    -
    1258  xt.setRangeStart(at_offset);
    -
    1259  xt.setHeight(nrefls_tile);
    -
    1260 
    -
    1261  comm::broadcast(rank_v0.col(), x, xt, mpi_row_chain, mpi_col_chain);
    -
    1262 
    -
    1263  // TRAILING MATRIX UPDATE
    -
    1264 
    -
    1265  // Note:
    -
    1266  // This trigger mechanism allows to control when the next iteration of compute panel will start.
    -
    1267  //
    -
    1268  // * What?
    -
    1269  // Compute panel uses MPI blocking communication that might block the only computing thread
    -
    1270  // available (since blocking communication are scheduled on normal queues and not on the MPI
    -
    1271  // dedicated one).
    -
    1272  //
    -
    1273  // * How?
    -
    1274  // If pika runtime has only 2 threads, one is dedicated to MPI and there is just one for
    -
    1275  // computation, that might get blocked by blocking MPI communication, without the chance to do
    -
    1276  // anything else. (TODO this might happen even with more reductions happening in parallel)
    -
    1277  //
    -
    1278  // * Why?
    -
    1279  // Panel computation at step i is done on the first column of the trailing matrix computed
    -
    1280  // at step i-1.
    -
    1281  // The rank owning the top-left tile of the trailing matrix, can update it as soon as it
    -
    1282  // receives X[0], which due to the pivot position is also the Xt[0]. Once it can go to the next
    -
    1283  // iteration, it ends up stucked in an MPI blocking communication, waiting for the others joining
    -
    1284  // before being able to advance.
    -
    1285  //
    -
    1286  // But at the same time, other ranks in the same column (needed for the next panel update), cannot
    -
    1287  // complete the trailing matrix update. Indeed, they are waiting for the pivot rank to communicate
    -
    1288  // column-wise Xt[0] (during x -> xt panel transpose broadcast), but he is not going to schedule
    -
    1289  // anything because the only normal thread which can do that is stuck in an MPI blocking
    -
    1290  // communication that is not going to advance... and so it's a DEADLOCK!
    -
    1291  //
    -
    1292  // * Solution:
    -
    1293  // The idea is to make the next panel depending not only on tiles stored locally, but also to
    -
    1294  // ensure that others have received Xt[0], which is needed to advance the computation and let
    -
    1295  // others arrive at the next iteration where the pivot will wait for them to complete the MPI
    -
    1296  // blocking communication.
    -
    1297  //
    -
    1298  // * Why is it different between MC and GPU?
    -
    1299  // As said above, the problem is related to the communication. But the communication is not said
    -
    1300  // to be an atomic operation happening in a single task. It might have to create a copy to
    -
    1301  // a buffer more suitable for the communication (e.g. GPU -> CPU if RDMA is not available).
    -
    1302  //
    -
    1303  // And in order to not be blocked, it must be ensured that the actual communication task has
    -
    1304  // been scheduled.
    -
    1305  const SizeType j_tile_current = ij_offset.col() / dist.blockSize().cols();
    -
    1306  const SizeType j_tile_next = at_offset.col() / dist.blockSize().cols();
    -
    1307  const bool isNextColumnOnSameRank = (j_tile_current == j_tile_next);
    -
    1308  const comm::IndexT_MPI rank_next_col =
    -
    1309  isNextColumnOnSameRank ? rank_v0.col() : (rank_v0.col() + 1) % dist.commGridSize().cols();
    -
    1310 
    -
    1311  if (rank.col() == rank_next_col) {
    -
    1312  const LocalTileIndex at{
    -
    1313  dist.template nextLocalTileFromGlobalElement<Coord::Row>(at_offset.row()),
    -
    1314  dist.template nextLocalTileFromGlobalElement<Coord::Col>(at_offset.col()),
    -
    1315  };
    -
    1316 
    -
    1317  // Note:
    -
    1318  // This additional communication of the last tile is a workaround for supporting following trigger
    -
    1319  // when b < mb.
    -
    1320  // Indeed, if b < mb the last column have (at least) a panel to compute, but differently from
    -
    1321  // other columns, broadcast transposed doesn't communicate the last tile, which is an assumption
    -
    1322  // needed to make the following trigger work correctly.
    -
    1323  const SizeType at_tile_col =
    -
    1324  dist.template globalTileFromGlobalElement<Coord::Col>(at_offset.col());
    -
    1325 
    -
    1326  if (at_tile_col == dist.nrTiles().cols() - 1) {
    -
    1327  const comm::IndexT_MPI owner = rank_v0.row();
    -
    1328  if (rank.row() == owner) {
    -
    1329  xt.setTile(at, x.read(at));
    -
    1330 
    -
    1331  if (dist.commGridSize().rows() > 1)
    -
    1332  ex::start_detached(comm::scheduleSendBcast(mpi_col_chain(), xt.read(at)));
    -
    1333  }
    -
    1334  else {
    -
    1335  if (dist.commGridSize().rows() > 1)
    -
    1336  ex::start_detached(comm::scheduleRecvBcast(mpi_col_chain(), owner, xt.readwrite(at)));
    -
    1337  }
    -
    1338  }
    -
    1339 
    -
    1340  if constexpr (dlaf::comm::CommunicationDevice_v<D> == D) {
    -
    1341  // Note:
    -
    1342  // if there is no need for additional buffers, we can just wait that xt[0] is ready for
    -
    1343  // reading.
    -
    1344  if (rank.row() == rank_v0.row()) {
    -
    1345  trigger_panel = xt.read(at) | ex::drop_value() | ex::ensure_started();
    -
    1346  }
    -
    1347  else {
    -
    1348  // Note:
    -
    1349  // Conservatively ensure that xt[0] needed for updating the first column has been
    -
    1350  // received. Just wait for xt because communication of x happens over rows, while the
    -
    1351  // pivot rank can just block rank in the same column.
    -
    1352  trigger_panel = xt.read(at) | ex::drop_value() | ex::ensure_started();
    -
    1353  }
    -
    1354  }
    -
    1355  else {
    -
    1356  if (rank.row() == rank_v0.row()) {
    -
    1357  // Note:
    -
    1358  // on the pivot rank, i.e. the one that would quickly go to the next panel and block, from
    -
    1359  // implementation we know that xt[0] is set as an external tile pointing to x[0].
    -
    1360  // We cannot wait on xt readwrite (because it is an external tile in a panel, that constraints
    -
    1361  // it to be just readable), but we can wait on its source x[0]. This has a subtle implication,
    -
    1362  // since we will wait not just for the communication to be complete (which is already more
    -
    1363  // than what needed), but we will also wait till xt[0] will be released, so after all local
    -
    1364  // communication and computation on the first column of the trailing matrix will be completed.
    -
    1365  trigger_panel = x.readwrite(at) | ex::drop_value() | ex::ensure_started();
    -
    1366  }
    -
    1367  else {
    -
    1368  // Note:
    -
    1369  // Conservatively ensure that xt[0] needed for updating the first column has been
    -
    1370  // received. Just wait for xt because communication of x happens over rows, while the
    -
    1371  // pivot rank can just block rank in the same column.
    -
    1372  trigger_panel = xt.read(at) | ex::drop_value() | ex::ensure_started();
    -
    1373  }
    -
    1374  }
    -
    1375  }
    -
    1376 
    -
    1377  // At -= X . V* + V . X*
    -
    1378  her2kUpdateTrailingMatrix<B>(trailing_matrix_view, mat_a, x, vt, v, xt);
    -
    1379 
    -
    1380  xt.reset();
    -
    1381  x.reset();
    -
    1382  wt.reset();
    -
    1383  w.reset();
    -
    1384  vt.reset();
    -
    1385  v.reset();
    -
    1386  }
    -
    1387 
    -
    1388  return mat_taus;
    +
    823 
    +
    824 template <Backend B, Device D, class T>
    +
    825 struct ComputePanelHelper;
    +
    826 
    +
    827 template <class T>
    +
    828 struct ComputePanelHelper<Backend::MC, Device::CPU, T> {
    +
    829  ComputePanelHelper(const std::size_t, matrix::Distribution) {}
    +
    830 
    +
    831  void call(Matrix<T, Device::CPU>& mat_a, Matrix<T, Device::CPU>& mat_taus, const SizeType j_sub,
    +
    832  const matrix::SubPanelView& panel_view) {
    +
    833  using red2band::local::computePanelReflectors;
    +
    834  computePanelReflectors(mat_a, mat_taus, j_sub, panel_view);
    +
    835  }
    +
    836 
    +
    837  template <Device D, class CommSender, class TriggerSender>
    +
    838  void call(TriggerSender&& trigger, comm::IndexT_MPI rank_v0, CommSender&& mpi_col_chain_panel,
    +
    839  Matrix<T, D>& mat_a, Matrix<T, Device::CPU>& mat_taus, const SizeType j_sub,
    +
    840  const matrix::SubPanelView& panel_view) {
    +
    841  using red2band::distributed::computePanelReflectors;
    +
    842  computePanelReflectors(std::forward<TriggerSender>(trigger), rank_v0,
    +
    843  std::forward<CommSender>(mpi_col_chain_panel), mat_a, mat_taus, j_sub,
    +
    844  panel_view);
    +
    845  }
    +
    846 };
    +
    847 
    +
    848 #ifdef DLAF_WITH_GPU
    +
    849 template <class T>
    +
    850 struct ComputePanelHelper<Backend::GPU, Device::GPU, T> {
    +
    851  ComputePanelHelper(const std::size_t n_workspaces, matrix::Distribution dist_a)
    +
    852  : panels_v(n_workspaces, dist_a) {}
    +
    853 
    +
    854  void call(Matrix<T, Device::GPU>& mat_a, Matrix<T, Device::CPU>& mat_taus, const SizeType j_sub,
    +
    855  const matrix::SubPanelView& panel_view) {
    +
    856  using red2band::local::computePanelReflectors;
    +
    857 
    +
    858  namespace ex = pika::execution::experimental;
    +
    859 
    +
    860  // Note:
    +
    861  // - copy panel_view from GPU to CPU
    +
    862  // - computePanelReflectors on CPU (on a matrix like, with just a panel)
    +
    863  // - copy back matrix "panel" from CPU to GPU
    +
    864 
    +
    865  auto& v = panels_v.nextResource();
    +
    866 
    +
    867  copyToCPU(panel_view, mat_a, v);
    +
    868  computePanelReflectors(v, mat_taus, j_sub, panel_view);
    +
    869  copyFromCPU(panel_view, v, mat_a);
    +
    870  }
    +
    871 
    +
    872  template <Device D, class CommSender, class TriggerSender>
    +
    873  void call(TriggerSender&& trigger, comm::IndexT_MPI rank_v0, CommSender&& mpi_col_chain_panel,
    +
    874  Matrix<T, D>& mat_a, Matrix<T, Device::CPU>& mat_taus, SizeType j_sub,
    +
    875  const matrix::SubPanelView& panel_view) {
    +
    876  auto& v = panels_v.nextResource();
    +
    877 
    +
    878  // copy to CPU
    +
    879  copyToCPU(panel_view, mat_a, v);
    +
    880 
    +
    881  // compute on CPU
    +
    882  using dlaf::eigensolver::internal::red2band::distributed::computePanelReflectors;
    +
    883  computePanelReflectors(std::forward<TriggerSender>(trigger), rank_v0,
    +
    884  std::forward<CommSender>(mpi_col_chain_panel), v, mat_taus, j_sub,
    +
    885  panel_view);
    +
    886 
    +
    887  // copy back to GPU
    +
    888  copyFromCPU(panel_view, v, mat_a);
    +
    889  }
    +
    890 
    +
    891 protected:
    +
    892  common::RoundRobin<matrix::Panel<Coord::Col, T, Device::CPU>> panels_v;
    +
    893 
    +
    894  void copyToCPU(const matrix::SubPanelView panel_view, matrix::Matrix<T, Device::GPU>& mat_a,
    +
    895  matrix::Panel<Coord::Col, T, Device::CPU>& v) {
    +
    896  namespace ex = pika::execution::experimental;
    +
    897 
    +
    898  using dlaf::internal::Policy;
    +
    899  using dlaf::matrix::internal::CopyBackend_v;
    +
    900  using pika::execution::thread_priority;
    +
    901 
    +
    902  for (const auto& i : panel_view.iteratorLocal()) {
    +
    903  auto spec = panel_view(i);
    +
    904  auto tmp_tile = v.readwrite(i);
    +
    905  ex::start_detached(
    +
    906  ex::when_all(splitTile(mat_a.read(i), spec), splitTile(std::move(tmp_tile), spec)) |
    +
    907  matrix::copy(Policy<CopyBackend_v<Device::GPU, Device::CPU>>(thread_priority::high)));
    +
    908  }
    +
    909  }
    +
    910 
    +
    911  void copyFromCPU(const matrix::SubPanelView panel_view, matrix::Panel<Coord::Col, T, Device::CPU>& v,
    +
    912  matrix::Matrix<T, Device::GPU>& mat_a) {
    +
    913  namespace ex = pika::execution::experimental;
    +
    914 
    +
    915  using dlaf::internal::Policy;
    +
    916  using dlaf::matrix::internal::CopyBackend_v;
    +
    917  using pika::execution::thread_priority;
    +
    918 
    +
    919  for (const auto& i : panel_view.iteratorLocal()) {
    +
    920  auto spec = panel_view(i);
    +
    921  auto tile_a = mat_a.readwrite(i);
    +
    922  ex::start_detached(
    +
    923  ex::when_all(splitTile(v.read(i), spec), splitTile(std::move(tile_a), spec)) |
    +
    924  matrix::copy(Policy<CopyBackend_v<Device::CPU, Device::GPU>>(thread_priority::high)));
    +
    925  }
    +
    926  }
    +
    927 };
    +
    928 #endif
    +
    929 
    +
    930 }
    +
    931 
    +
    932 // Local implementation of reduction to band
    +
    933 template <Backend B, Device D, class T>
    +
    934 Matrix<T, Device::CPU> ReductionToBand<B, D, T>::call(Matrix<T, D>& mat_a, const SizeType band_size) {
    +
    935  using dlaf::matrix::Matrix;
    +
    936  using dlaf::matrix::Panel;
    +
    937 
    +
    938  using namespace red2band::local;
    +
    939 
    +
    940  using common::iterate_range2d;
    +
    941  using factorization::internal::computeTFactor;
    +
    942 
    +
    943  using pika::execution::experimental::any_sender;
    +
    944 
    +
    945  const auto dist_a = mat_a.distribution();
    +
    946  const matrix::Distribution dist({mat_a.size().rows(), band_size},
    +
    947  {dist_a.blockSize().rows(), band_size});
    +
    948 
    +
    949  // Note:
    +
    950  // Reflector of size = 1 is not considered whatever T is (i.e. neither real nor complex)
    +
    951  const SizeType nrefls = std::max<SizeType>(0, dist_a.size().rows() - band_size - 1);
    +
    952 
    +
    953  // Row-vector that is distributed over columns, but exists locally on all rows of the grid
    +
    954  DLAF_ASSERT(mat_a.blockSize().cols() % band_size == 0, mat_a.blockSize().cols(), band_size);
    +
    955  Matrix<T, Device::CPU> mat_taus(matrix::Distribution(GlobalElementSize(nrefls, 1),
    +
    956  TileElementSize(mat_a.blockSize().cols(), 1),
    +
    957  comm::Size2D(mat_a.commGridSize().cols(), 1),
    +
    958  comm::Index2D(mat_a.rankIndex().col(), 0),
    +
    959  comm::Index2D(mat_a.sourceRankIndex().col(), 0)));
    +
    960 
    +
    961  if (nrefls == 0)
    +
    962  return mat_taus;
    +
    963 
    +
    964  Matrix<T, Device::CPU> mat_taus_retiled =
    +
    965  mat_taus.retiledSubPipeline(LocalTileSize(mat_a.blockSize().cols() / band_size, 1));
    +
    966 
    +
    967  const SizeType ntiles = (nrefls - 1) / band_size + 1;
    +
    968  DLAF_ASSERT(ntiles == mat_taus_retiled.nrTiles().rows(), ntiles, mat_taus_retiled.nrTiles().rows());
    +
    969 
    +
    970  const bool is_full_band = (band_size == dist_a.blockSize().cols());
    +
    971 
    +
    972  constexpr std::size_t n_workspaces = 2;
    +
    973  common::RoundRobin<Panel<Coord::Col, T, D>> panels_v(n_workspaces, dist);
    +
    974  common::RoundRobin<Panel<Coord::Col, T, D>> panels_w(n_workspaces, dist);
    +
    975  common::RoundRobin<Panel<Coord::Col, T, D>> panels_x(n_workspaces, dist);
    +
    976 
    +
    977  // Note:
    +
    978  // Here dist_a is given with full panel size instead of dist with just the part actually needeed,
    +
    979  // because the GPU Helper internally exploits Panel data-structure. Indeed, the full size panel is
    +
    980  // needed in order to mimick Matrix with Panel, so it is possible to apply a SubPanelView to it.
    +
    981  //
    +
    982  // It is a bit hacky usage, because SubPanelView is not meant to be used with Panel, but just with
    +
    983  // Matrix. This results in a variable waste of memory, depending no the ratio band_size/nb.
    +
    984  red2band::ComputePanelHelper<B, D, T> compute_panel_helper(n_workspaces, dist_a);
    +
    985 
    +
    986  for (SizeType j_sub = 0; j_sub < ntiles; ++j_sub) {
    +
    987  const auto i_sub = j_sub + 1;
    +
    988 
    +
    989  const GlobalElementIndex ij_offset(i_sub * band_size, j_sub * band_size);
    +
    990 
    +
    991  const SizeType nrefls_tile = mat_taus_retiled.tileSize(GlobalTileIndex(j_sub, 0)).rows();
    +
    992 
    +
    993  const bool isPanelIncomplete = (nrefls_tile != band_size);
    +
    994 
    +
    995  // Note: if this is running, it must have at least one valid reflector (i.e. with size > 1)
    +
    996  DLAF_ASSERT_HEAVY(nrefls_tile != 0, nrefls_tile);
    +
    997 
    +
    998  // Note: SubPanelView is (at most) band_size wide, but it may contain a smaller number of
    +
    999  // reflectors (i.e. at the end when last reflector size is 1)
    +
    1000  const matrix::SubPanelView panel_view(dist_a, ij_offset, band_size);
    +
    1001 
    +
    1002  Panel<Coord::Col, T, D>& v = panels_v.nextResource();
    +
    1003  v.setRangeStart(ij_offset);
    +
    1004  if (isPanelIncomplete)
    +
    1005  v.setWidth(nrefls_tile);
    +
    1006 
    +
    1007  // PANEL
    +
    1008  compute_panel_helper.call(mat_a, mat_taus_retiled, j_sub, panel_view);
    +
    1009 
    +
    1010  // Note:
    +
    1011  // - has_reflector_head tells if this rank owns the first tile of the panel (being local, always true)
    +
    1012  // - if !is_full_band it has to force copy as a workaround, otherwise in update matrix it would deadlock
    +
    1013  // due to tile shared between panel and trailing matrix
    +
    1014  constexpr bool has_reflector_head = true;
    +
    1015  setupReflectorPanelV<B, D, T>(has_reflector_head, panel_view, nrefls_tile, v, mat_a, !is_full_band);
    +
    1016 
    +
    1017  const LocalTileIndex t_idx(0, 0);
    +
    1018  // TODO used just by the column, maybe we can re-use a panel tile?
    +
    1019  // TODO probably the first one in any panel is ok?
    +
    1020  Matrix<T, D> t({nrefls_tile, nrefls_tile}, dist.blockSize());
    +
    1021 
    +
    1022  computeTFactor<B>(v, mat_taus_retiled.read(GlobalTileIndex(j_sub, 0)), t.readwrite(t_idx));
    +
    1023 
    +
    1024  // PREPARATION FOR TRAILING MATRIX UPDATE
    +
    1025  const GlobalElementIndex at_offset(ij_offset + GlobalElementSize(0, band_size));
    +
    1026 
    +
    1027  // Note: if there is no trailing matrix, algorithm has finised
    +
    1028  if (!at_offset.isIn(mat_a.size()))
    +
    1029  break;
    +
    1030 
    +
    1031  const matrix::SubMatrixView trailing_matrix_view(dist_a, at_offset);
    +
    1032 
    +
    1033  // W = V . T
    +
    1034  Panel<Coord::Col, T, D>& w = panels_w.nextResource();
    +
    1035  w.setRangeStart(at_offset);
    +
    1036  if (isPanelIncomplete)
    +
    1037  w.setWidth(nrefls_tile);
    +
    1038 
    +
    1039  trmmComputeW<B>(w, v, t.read(t_idx));
    +
    1040 
    +
    1041  // X = At . W
    +
    1042  Panel<Coord::Col, T, D>& x = panels_x.nextResource();
    +
    1043  x.setRangeStart(at_offset);
    +
    1044  if (isPanelIncomplete)
    +
    1045  x.setWidth(nrefls_tile);
    +
    1046 
    +
    1047  // Note:
    +
    1048  // Since At is hermitian, just the lower part is referenced.
    +
    1049  // When the tile is not part of the main diagonal, the same tile has to be used for two computations
    +
    1050  // that will contribute to two different rows of X: the ones indexed with row and col.
    +
    1051  hemmComputeX<B>(x, trailing_matrix_view, mat_a, w);
    +
    1052 
    +
    1053  // In the next section the next two operations are performed
    +
    1054  // A) W2 = W* . X
    +
    1055  // B) X -= 1/2 . V . W2
    +
    1056 
    +
    1057  // Note:
    +
    1058  // T can be re-used because it is not needed anymore in this step and it has the same shape
    +
    1059  Matrix<T, D> w2 = std::move(t);
    +
    1060 
    +
    1061  gemmComputeW2<B>(w2, w, x);
    +
    1062  gemmUpdateX<B>(x, w2, v);
    +
    1063 
    +
    1064  // TRAILING MATRIX UPDATE
    +
    1065 
    +
    1066  // At -= X . V* + V . X*
    +
    1067  her2kUpdateTrailingMatrix<B>(trailing_matrix_view, mat_a, x, v);
    +
    1068 
    +
    1069  x.reset();
    +
    1070  w.reset();
    +
    1071  v.reset();
    +
    1072  }
    +
    1073 
    +
    1074  return mat_taus;
    +
    1075 }
    +
    1076 
    +
    1077 // Distributed implementation of reduction to band
    +
    1078 template <Backend B, Device D, class T>
    +
    1079 Matrix<T, Device::CPU> ReductionToBand<B, D, T>::call(comm::CommunicatorGrid grid, Matrix<T, D>& mat_a,
    +
    1080  const SizeType band_size) {
    +
    1081  using namespace red2band::distributed;
    +
    1082 
    +
    1083  using common::iterate_range2d;
    +
    1084  using factorization::internal::computeTFactor;
    +
    1085 
    +
    1086  namespace ex = pika::execution::experimental;
    +
    1087 
    +
    1088  // Note:
    +
    1089  // This is a temporary workaround.
    +
    1090  // See issue https://github.com/eth-cscs/DLA-Future/issues/729
    +
    1091  pika::threads::get_thread_manager().wait();
    +
    1092 
    +
    1093  common::Pipeline<comm::Communicator> mpi_col_chain_panel(grid.colCommunicator().clone());
    +
    1094  common::Pipeline<comm::Communicator> mpi_row_chain(grid.rowCommunicator().clone());
    +
    1095  common::Pipeline<comm::Communicator> mpi_col_chain(grid.colCommunicator().clone());
    +
    1096 
    +
    1097  const auto& dist = mat_a.distribution();
    +
    1098  const comm::Index2D rank = dist.rankIndex();
    +
    1099 
    +
    1100  // Note:
    +
    1101  // Reflector of size = 1 is not considered whatever T is (i.e. neither real nor complex)
    +
    1102  const SizeType nrefls = std::max<SizeType>(0, dist.size().rows() - band_size - 1);
    +
    1103 
    +
    1104  // Row-vector that is distributed over columns, but exists locally on all rows of the grid
    +
    1105  DLAF_ASSERT(mat_a.blockSize().cols() % band_size == 0, mat_a.blockSize().cols(), band_size);
    +
    1106  Matrix<T, Device::CPU> mat_taus(matrix::Distribution(GlobalElementSize(nrefls, 1),
    +
    1107  TileElementSize(mat_a.blockSize().cols(), 1),
    +
    1108  comm::Size2D(mat_a.commGridSize().cols(), 1),
    +
    1109  comm::Index2D(mat_a.rankIndex().col(), 0),
    +
    1110  comm::Index2D(mat_a.sourceRankIndex().col(), 0)));
    +
    1111 
    +
    1112  if (nrefls == 0)
    +
    1113  return mat_taus;
    +
    1114 
    +
    1115  Matrix<T, Device::CPU> mat_taus_retiled =
    +
    1116  mat_taus.retiledSubPipeline(LocalTileSize(mat_a.blockSize().cols() / band_size, 1));
    +
    1117 
    +
    1118  const SizeType ntiles = (nrefls - 1) / band_size + 1;
    +
    1119  DLAF_ASSERT(ntiles == mat_taus_retiled.nrTiles().rows(), ntiles, mat_taus_retiled.nrTiles().rows());
    +
    1120 
    +
    1121  const bool is_full_band = (band_size == dist.blockSize().cols());
    +
    1122 
    +
    1123  constexpr std::size_t n_workspaces = 2;
    +
    1124  common::RoundRobin<matrix::Panel<Coord::Col, T, D>> panels_v(n_workspaces, dist);
    +
    1125  common::RoundRobin<matrix::Panel<Coord::Row, T, D, matrix::StoreTransposed::Yes>> panels_vt(
    +
    1126  n_workspaces, dist);
    +
    1127 
    +
    1128  common::RoundRobin<matrix::Panel<Coord::Col, T, D>> panels_w(n_workspaces, dist);
    +
    1129  common::RoundRobin<matrix::Panel<Coord::Row, T, D, matrix::StoreTransposed::Yes>> panels_wt(
    +
    1130  n_workspaces, dist);
    +
    1131 
    +
    1132  common::RoundRobin<matrix::Panel<Coord::Col, T, D>> panels_x(n_workspaces, dist);
    +
    1133  common::RoundRobin<matrix::Panel<Coord::Row, T, D, matrix::StoreTransposed::Yes>> panels_xt(
    +
    1134  n_workspaces, dist);
    +
    1135 
    +
    1136  red2band::ComputePanelHelper<B, D, T> compute_panel_helper(n_workspaces, dist);
    +
    1137 
    +
    1138  ex::unique_any_sender<> trigger_panel{ex::just()};
    +
    1139  for (SizeType j_sub = 0; j_sub < ntiles; ++j_sub) {
    +
    1140  const SizeType i_sub = j_sub + 1;
    +
    1141 
    +
    1142  const GlobalElementIndex ij_offset(i_sub * band_size, j_sub * band_size);
    +
    1143  const GlobalElementIndex at_offset(i_sub * band_size, (j_sub + 1) * band_size);
    +
    1144 
    +
    1145  const comm::Index2D rank_v0{
    +
    1146  dist.template rankGlobalElement<Coord::Row>(ij_offset.row()),
    +
    1147  dist.template rankGlobalElement<Coord::Col>(ij_offset.col()),
    +
    1148  };
    +
    1149 
    +
    1150  const bool is_panel_rank_col = rank_v0.col() == rank.col();
    +
    1151 
    +
    1152  const SizeType nrefls_tile = mat_taus_retiled.tileSize(GlobalTileIndex(j_sub, 0)).rows();
    +
    1153 
    +
    1154  if (nrefls_tile == 0)
    +
    1155  break;
    +
    1156 
    +
    1157  auto& v = panels_v.nextResource();
    +
    1158  auto& vt = panels_vt.nextResource();
    +
    1159 
    +
    1160  v.setRangeStart(at_offset);
    +
    1161  vt.setRangeStart(at_offset);
    +
    1162 
    +
    1163  v.setWidth(nrefls_tile);
    +
    1164  vt.setHeight(nrefls_tile);
    +
    1165 
    +
    1166  const LocalTileIndex t_idx(0, 0);
    +
    1167  // TODO used just by the column, maybe we can re-use a panel tile?
    +
    1168  // TODO or we can keep just the sh_future and allocate just inside if (is_panel_rank_col)
    +
    1169  matrix::Matrix<T, D> t({nrefls_tile, nrefls_tile}, dist.blockSize());
    +
    1170 
    +
    1171  // PANEL
    +
    1172  const matrix::SubPanelView panel_view(dist, ij_offset, band_size);
    +
    1173 
    +
    1174  if (is_panel_rank_col) {
    +
    1175  compute_panel_helper.call(std::move(trigger_panel), rank_v0.row(), mpi_col_chain_panel(), mat_a,
    +
    1176  mat_taus_retiled, j_sub, panel_view);
    +
    1177 
    +
    1178  // Note:
    +
    1179  // - has_reflector_head tells if this rank owns the first tile of the panel
    +
    1180  // - if !is_full_band it has to force copy as a workaround, otherwise in update matrix it would
    +
    1181  // deadlock due to tile shared between panel and trailing matrix
    +
    1182  red2band::local::setupReflectorPanelV<B, D, T>(rank.row() == rank_v0.row(), panel_view,
    +
    1183  nrefls_tile, v, mat_a, !is_full_band);
    +
    1184  computeTFactor<B>(v, mat_taus_retiled.read(GlobalTileIndex(j_sub, 0)), t.readwrite(t_idx),
    +
    1185  mpi_col_chain);
    +
    1186  }
    +
    1187 
    +
    1188  // PREPARATION FOR TRAILING MATRIX UPDATE
    +
    1189 
    +
    1190  // Note: if there is no trailing matrix, algorithm has finised
    +
    1191  if (!at_offset.isIn(mat_a.size()))
    +
    1192  break;
    +
    1193 
    +
    1194  const matrix::SubMatrixView trailing_matrix_view(dist, at_offset);
    +
    1195 
    +
    1196  comm::broadcast(rank_v0.col(), v, vt, mpi_row_chain, mpi_col_chain);
    +
    1197 
    +
    1198  // W = V . T
    +
    1199  auto& w = panels_w.nextResource();
    +
    1200  auto& wt = panels_wt.nextResource();
    +
    1201 
    +
    1202  w.setRangeStart(at_offset);
    +
    1203  wt.setRangeStart(at_offset);
    +
    1204 
    +
    1205  w.setWidth(nrefls_tile);
    +
    1206  wt.setHeight(nrefls_tile);
    +
    1207 
    +
    1208  if (is_panel_rank_col)
    +
    1209  red2band::local::trmmComputeW<B, D>(w, v, t.read(t_idx));
    +
    1210 
    +
    1211  comm::broadcast(rank_v0.col(), w, wt, mpi_row_chain, mpi_col_chain);
    +
    1212 
    +
    1213  // X = At . W
    +
    1214  auto& x = panels_x.nextResource();
    +
    1215  auto& xt = panels_xt.nextResource();
    +
    1216 
    +
    1217  x.setRangeStart(at_offset);
    +
    1218  xt.setRangeStart(at_offset);
    +
    1219 
    +
    1220  x.setWidth(nrefls_tile);
    +
    1221  xt.setHeight(nrefls_tile);
    +
    1222 
    +
    1223  // Note:
    +
    1224  // Since At is hermitian, just the lower part is referenced.
    +
    1225  // When the tile is not part of the main diagonal, the same tile has to be used for two computations
    +
    1226  // that will contribute to two different rows of X: the ones indexed with row and col.
    +
    1227  // This is achieved by storing the two results in two different workspaces: X and X_conj respectively.
    +
    1228  //
    +
    1229  // On exit, x will contain a valid result just on ranks belonging to the column panel.
    +
    1230  // For what concerns xt, it is just used as support and it contains junk data on all ranks.
    +
    1231  hemmComputeX<B, D>(rank_v0.col(), x, xt, trailing_matrix_view, mat_a, w, wt, mpi_row_chain,
    +
    1232  mpi_col_chain);
    +
    1233 
    +
    1234  // In the next section the next two operations are performed
    +
    1235  // A) W2 = W* . X
    +
    1236  // B) X -= 1/2 . V . W2
    +
    1237 
    +
    1238  // Note:
    +
    1239  // Now the intermediate result for X is available on the panel column ranks,
    +
    1240  // which have locally all the needed stuff for updating X and finalize the result
    +
    1241  if (is_panel_rank_col) {
    +
    1242  // Note:
    +
    1243  // T can be re-used because it is not needed anymore in this step and it has the same shape
    +
    1244  matrix::Matrix<T, D> w2 = std::move(t);
    +
    1245 
    +
    1246  red2band::local::gemmComputeW2<B, D>(w2, w, x);
    +
    1247  ex::start_detached(comm::scheduleAllReduceInPlace(mpi_col_chain(), MPI_SUM,
    +
    1248  w2.readwrite(LocalTileIndex(0, 0))));
    +
    1249 
    +
    1250  red2band::local::gemmUpdateX<B, D>(x, w2, v);
    +
    1251  }
    +
    1252 
    +
    1253  // Note:
    +
    1254  // xt has been used previously as workspace for hemmComputeX, so it has to be reset, because now it
    +
    1255  // will be used for accessing the broadcasted version of x
    +
    1256  xt.reset();
    +
    1257  xt.setRangeStart(at_offset);
    +
    1258  xt.setHeight(nrefls_tile);
    +
    1259 
    +
    1260  comm::broadcast(rank_v0.col(), x, xt, mpi_row_chain, mpi_col_chain);
    +
    1261 
    +
    1262  // TRAILING MATRIX UPDATE
    +
    1263 
    +
    1264  // Note:
    +
    1265  // This trigger mechanism allows to control when the next iteration of compute panel will start.
    +
    1266  //
    +
    1267  // * What?
    +
    1268  // Compute panel uses MPI blocking communication that might block the only computing thread
    +
    1269  // available (since blocking communication are scheduled on normal queues and not on the MPI
    +
    1270  // dedicated one).
    +
    1271  //
    +
    1272  // * How?
    +
    1273  // If pika runtime has only 2 threads, one is dedicated to MPI and there is just one for
    +
    1274  // computation, that might get blocked by blocking MPI communication, without the chance to do
    +
    1275  // anything else. (TODO this might happen even with more reductions happening in parallel)
    +
    1276  //
    +
    1277  // * Why?
    +
    1278  // Panel computation at step i is done on the first column of the trailing matrix computed
    +
    1279  // at step i-1.
    +
    1280  // The rank owning the top-left tile of the trailing matrix, can update it as soon as it
    +
    1281  // receives X[0], which due to the pivot position is also the Xt[0]. Once it can go to the next
    +
    1282  // iteration, it ends up stucked in an MPI blocking communication, waiting for the others joining
    +
    1283  // before being able to advance.
    +
    1284  //
    +
    1285  // But at the same time, other ranks in the same column (needed for the next panel update), cannot
    +
    1286  // complete the trailing matrix update. Indeed, they are waiting for the pivot rank to communicate
    +
    1287  // column-wise Xt[0] (during x -> xt panel transpose broadcast), but he is not going to schedule
    +
    1288  // anything because the only normal thread which can do that is stuck in an MPI blocking
    +
    1289  // communication that is not going to advance... and so it's a DEADLOCK!
    +
    1290  //
    +
    1291  // * Solution:
    +
    1292  // The idea is to make the next panel depending not only on tiles stored locally, but also to
    +
    1293  // ensure that others have received Xt[0], which is needed to advance the computation and let
    +
    1294  // others arrive at the next iteration where the pivot will wait for them to complete the MPI
    +
    1295  // blocking communication.
    +
    1296  //
    +
    1297  // * Why is it different between MC and GPU?
    +
    1298  // As said above, the problem is related to the communication. But the communication is not said
    +
    1299  // to be an atomic operation happening in a single task. It might have to create a copy to
    +
    1300  // a buffer more suitable for the communication (e.g. GPU -> CPU if RDMA is not available).
    +
    1301  //
    +
    1302  // And in order to not be blocked, it must be ensured that the actual communication task has
    +
    1303  // been scheduled.
    +
    1304  const SizeType j_tile_current = ij_offset.col() / dist.blockSize().cols();
    +
    1305  const SizeType j_tile_next = at_offset.col() / dist.blockSize().cols();
    +
    1306  const bool isNextColumnOnSameRank = (j_tile_current == j_tile_next);
    +
    1307  const comm::IndexT_MPI rank_next_col =
    +
    1308  isNextColumnOnSameRank ? rank_v0.col() : (rank_v0.col() + 1) % dist.commGridSize().cols();
    +
    1309 
    +
    1310  if (rank.col() == rank_next_col) {
    +
    1311  const LocalTileIndex at{
    +
    1312  dist.template nextLocalTileFromGlobalElement<Coord::Row>(at_offset.row()),
    +
    1313  dist.template nextLocalTileFromGlobalElement<Coord::Col>(at_offset.col()),
    +
    1314  };
    +
    1315 
    +
    1316  // Note:
    +
    1317  // This additional communication of the last tile is a workaround for supporting following trigger
    +
    1318  // when b < mb.
    +
    1319  // Indeed, if b < mb the last column have (at least) a panel to compute, but differently from
    +
    1320  // other columns, broadcast transposed doesn't communicate the last tile, which is an assumption
    +
    1321  // needed to make the following trigger work correctly.
    +
    1322  const SizeType at_tile_col =
    +
    1323  dist.template globalTileFromGlobalElement<Coord::Col>(at_offset.col());
    +
    1324 
    +
    1325  if (at_tile_col == dist.nrTiles().cols() - 1) {
    +
    1326  const comm::IndexT_MPI owner = rank_v0.row();
    +
    1327  if (rank.row() == owner) {
    +
    1328  xt.setTile(at, x.read(at));
    +
    1329 
    +
    1330  if (dist.commGridSize().rows() > 1)
    +
    1331  ex::start_detached(comm::scheduleSendBcast(mpi_col_chain(), xt.read(at)));
    +
    1332  }
    +
    1333  else {
    +
    1334  if (dist.commGridSize().rows() > 1)
    +
    1335  ex::start_detached(comm::scheduleRecvBcast(mpi_col_chain(), owner, xt.readwrite(at)));
    +
    1336  }
    +
    1337  }
    +
    1338 
    +
    1339  if constexpr (dlaf::comm::CommunicationDevice_v<D> == D) {
    +
    1340  // Note:
    +
    1341  // if there is no need for additional buffers, we can just wait that xt[0] is ready for
    +
    1342  // reading.
    +
    1343  if (rank.row() == rank_v0.row()) {
    +
    1344  trigger_panel = xt.read(at) | ex::drop_value() | ex::ensure_started();
    +
    1345  }
    +
    1346  else {
    +
    1347  // Note:
    +
    1348  // Conservatively ensure that xt[0] needed for updating the first column has been
    +
    1349  // received. Just wait for xt because communication of x happens over rows, while the
    +
    1350  // pivot rank can just block rank in the same column.
    +
    1351  trigger_panel = xt.read(at) | ex::drop_value() | ex::ensure_started();
    +
    1352  }
    +
    1353  }
    +
    1354  else {
    +
    1355  if (rank.row() == rank_v0.row()) {
    +
    1356  // Note:
    +
    1357  // on the pivot rank, i.e. the one that would quickly go to the next panel and block, from
    +
    1358  // implementation we know that xt[0] is set as an external tile pointing to x[0].
    +
    1359  // We cannot wait on xt readwrite (because it is an external tile in a panel, that constraints
    +
    1360  // it to be just readable), but we can wait on its source x[0]. This has a subtle implication,
    +
    1361  // since we will wait not just for the communication to be complete (which is already more
    +
    1362  // than what needed), but we will also wait till xt[0] will be released, so after all local
    +
    1363  // communication and computation on the first column of the trailing matrix will be completed.
    +
    1364  trigger_panel = x.readwrite(at) | ex::drop_value() | ex::ensure_started();
    +
    1365  }
    +
    1366  else {
    +
    1367  // Note:
    +
    1368  // Conservatively ensure that xt[0] needed for updating the first column has been
    +
    1369  // received. Just wait for xt because communication of x happens over rows, while the
    +
    1370  // pivot rank can just block rank in the same column.
    +
    1371  trigger_panel = xt.read(at) | ex::drop_value() | ex::ensure_started();
    +
    1372  }
    +
    1373  }
    +
    1374  }
    +
    1375 
    +
    1376  // At -= X . V* + V . X*
    +
    1377  her2kUpdateTrailingMatrix<B>(trailing_matrix_view, mat_a, x, vt, v, xt);
    +
    1378 
    +
    1379  xt.reset();
    +
    1380  x.reset();
    +
    1381  wt.reset();
    +
    1382  w.reset();
    +
    1383  vt.reset();
    +
    1384  v.reset();
    +
    1385  }
    +
    1386 
    +
    1387  return mat_taus;
    +
    1388 }
    1389 }
    -
    1390 }
    broadcast_panel.h
    dlaf::internal::Policy
    Definition: policy.h:24
    dlaf::matrix::Distribution
    Definition: distribution.h:27
    -
    dlaf::matrix::Matrix
    Definition: matrix.h:55
    -
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
    Definition: matrix.h:121
    -
    dlaf::matrix::RetiledMatrix
    Definition: retiled_matrix.h:36
    +
    dlaf::matrix::Matrix
    Definition: matrix.h:56
    +
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
    Definition: matrix.h:122
    dlaf::comm::IndexT_MPI
    int IndexT_MPI
    Type used for indexes in MPI API.
    Definition: communicator.h:23
    communicator_grid.h
    @@ -1484,19 +1482,18 @@
    dlaf::matrix::splitTile
    ReadOnlyTileSender< T, D > splitTile(ReadOnlyTileSender< T, D > tile, const SubTileSpec &spec)
    Definition: tile.h:507
    -
    dlaf::auxiliary::norm
    dlaf::BaseType< T > norm(comm::CommunicatorGrid grid, comm::Index2D rank, lapack::Norm norm_type, blas::Uplo uplo, Matrix< const T, device > &A)
    Definition: norm.h:43
    +
    dlaf::auxiliary::norm
    dlaf::BaseType< T > norm(comm::CommunicatorGrid grid, comm::Index2D rank, lapack::Norm norm_type, blas::Uplo uplo, Matrix< const T, device > &A)
    Definition: norm.h:44
    -
    retiled_matrix.h
    round_robin.h
    single_threaded_blas.h
    dlaf::common::RoundRobin
    Definition: round_robin.h:20
    - +
    dlaf::matrix::Panel
    Definition: panel.h:552
    dlaf::matrix::Panel::readwrite
    ReadWriteSenderType readwrite(LocalTileIndex index)
    Definition: panel.h:570
    dlaf::matrix::SubPanelView
    Definition: views.h:132
    diff --git a/master/files.html b/master/files.html index cdbba16a25..a9d003ade2 100644 --- a/master/files.html +++ b/master/files.html @@ -216,10 +216,9 @@ - - - - + + + diff --git a/master/functions_d.html b/master/functions_d.html index af9b81d1ca..2b1efff2e7 100644 --- a/master/functions_d.html +++ b/master/functions_d.html @@ -91,6 +91,9 @@

    - d -

    diff --git a/master/functions_func_d.html b/master/functions_func_d.html index 7da8334928..1ae3105610 100644 --- a/master/functions_func_d.html +++ b/master/functions_func_d.html @@ -81,6 +81,9 @@

    - d -

    diff --git a/master/functions_func_r.html b/master/functions_func_r.html index 348a38a440..eec6e1a6bc 100644 --- a/master/functions_func_r.html +++ b/master/functions_func_r.html @@ -97,13 +97,11 @@

    - r -

      : dlaf::matrix::internal::TilePipeline< T, D > , dlaf::matrix::Matrix< const T, D > , dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No > -, dlaf::matrix::RetiledMatrix< T, D >
    • readwrite() : dlaf::matrix::internal::TilePipeline< T, D > , dlaf::matrix::Matrix< T, D > , dlaf::matrix::Panel< axis, T, D, Storage > -, dlaf::matrix::RetiledMatrix< T, D >
    • readwrite_with_wrapper() : dlaf::matrix::internal::TilePipeline< T, D > @@ -113,8 +111,11 @@

      - r -

        , dlaf::matrix::internal::TilePipeline< T, D > , dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No > -
      • RetiledMatrix() -: dlaf::matrix::RetiledMatrix< T, D > +
      • retiledSubPipeline() +: dlaf::matrix::Matrix< T, D > +
        • +
      • retiledSubPipelineConst() +: dlaf::matrix::Matrix< const T, D >
      • rowCommunicator() : dlaf::comm::CommunicatorGrid diff --git a/master/functions_func_s.html b/master/functions_func_s.html index 617197a5e0..34786b10a4 100644 --- a/master/functions_func_s.html +++ b/master/functions_func_s.html @@ -102,6 +102,12 @@

        - s -

        • SubPanelView() : dlaf::matrix::SubPanelView
          • +
        • subPipeline() +: dlaf::matrix::Matrix< T, D > +
          • +
        • subPipelineConst() +: dlaf::matrix::Matrix< const T, D > +
        • subTileReference() : dlaf::matrix::Tile< T, D > , dlaf::matrix::Tile< const T, D > diff --git a/master/functions_func_w.html b/master/functions_func_w.html index 413bbe4aff..69d218dfcc 100644 --- a/master/functions_func_w.html +++ b/master/functions_func_w.html @@ -66,7 +66,6 @@

          - w -

          diff --git a/master/functions_r.html b/master/functions_r.html index 0ec72bd5c5..d2e68d5e82 100644 --- a/master/functions_r.html +++ b/master/functions_r.html @@ -97,13 +97,11 @@

          - r -

            : dlaf::matrix::internal::TilePipeline< T, D > , dlaf::matrix::Matrix< const T, D > , dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No > -, dlaf::matrix::RetiledMatrix< T, D >
          • readwrite() : dlaf::matrix::internal::TilePipeline< T, D > , dlaf::matrix::Matrix< T, D > , dlaf::matrix::Panel< axis, T, D, Storage > -, dlaf::matrix::RetiledMatrix< T, D >
          • readwrite_with_wrapper() : dlaf::matrix::internal::TilePipeline< T, D > @@ -113,8 +111,11 @@

            - r -

              , dlaf::matrix::internal::TilePipeline< T, D > , dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No > -
            • RetiledMatrix() -: dlaf::matrix::RetiledMatrix< T, D > +
            • retiledSubPipeline() +: dlaf::matrix::Matrix< T, D > +
              • +
            • retiledSubPipelineConst() +: dlaf::matrix::Matrix< const T, D >
            • rowCommunicator() : dlaf::comm::CommunicatorGrid diff --git a/master/functions_s.html b/master/functions_s.html index 1444582871..bc1d3042b3 100644 --- a/master/functions_s.html +++ b/master/functions_s.html @@ -108,6 +108,12 @@

              - s -

              • SubPanelView() : dlaf::matrix::SubPanelView
                • +
              • subPipeline() +: dlaf::matrix::Matrix< T, D > +
                • +
              • subPipelineConst() +: dlaf::matrix::Matrix< const T, D > +
              • subTileReference() : dlaf::matrix::Tile< T, D > , dlaf::matrix::Tile< const T, D > diff --git a/master/functions_w.html b/master/functions_w.html index 89f0ef8197..49eb922c52 100644 --- a/master/functions_w.html +++ b/master/functions_w.html @@ -66,7 +66,6 @@

                - w -

                diff --git a/master/gen__eigensolver_8h_source.html b/master/gen__eigensolver_8h_source.html index aa4eb0c628..e4eadf64fb 100644 --- a/master/gen__eigensolver_8h_source.html +++ b/master/gen__eigensolver_8h_source.html @@ -110,83 +110,91 @@
                58  eigenvectors);
                59  DLAF_ASSERT(eigenvectors.size() == mat_a.size(), eigenvectors, mat_a);
                60  DLAF_ASSERT(eigenvectors.blockSize() == mat_a.blockSize(), eigenvectors, mat_a);
                -
                61 
                -
                62  internal::GenEigensolver<B, D, T>::call(uplo, mat_a, mat_b, eigenvalues, eigenvectors);
                -
                63 }
                -
                64 
                -
                80 template <Backend B, Device D, class T>
                -
                81 EigensolverResult<T, D> genEigensolver(blas::Uplo uplo, Matrix<T, D>& mat_a, Matrix<T, D>& mat_b) {
                -
                82  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
                -
                83  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
                -
                84  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                -
                85  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                -
                86  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                -
                87  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                -
                88  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                -
                89  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                -
                90 
                -
                91  const SizeType size = mat_a.size().rows();
                -
                92 
                -
                93  matrix::Matrix<BaseType<T>, D> eigenvalues(LocalElementSize(size, 1),
                -
                94  TileElementSize(mat_a.blockSize().rows(), 1));
                -
                95  matrix::Matrix<T, D> eigenvectors(LocalElementSize(size, size), mat_a.blockSize());
                +
                61  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
                +
                62  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
                +
                63  DLAF_ASSERT(matrix::single_tile_per_block(eigenvalues), eigenvalues);
                +
                64  DLAF_ASSERT(matrix::single_tile_per_block(eigenvectors), eigenvectors);
                +
                65 
                +
                66  internal::GenEigensolver<B, D, T>::call(uplo, mat_a, mat_b, eigenvalues, eigenvectors);
                +
                67 }
                +
                68 
                +
                84 template <Backend B, Device D, class T>
                +
                85 EigensolverResult<T, D> genEigensolver(blas::Uplo uplo, Matrix<T, D>& mat_a, Matrix<T, D>& mat_b) {
                +
                86  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
                +
                87  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
                +
                88  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                +
                89  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                +
                90  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                +
                91  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                +
                92  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                +
                93  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                +
                94 
                +
                95  const SizeType size = mat_a.size().rows();
                96 
                -
                97  genEigensolver<B, D, T>(uplo, mat_a, mat_b, eigenvalues, eigenvectors);
                -
                98 
                -
                99  return {std::move(eigenvalues), std::move(eigenvectors)};
                -
                100 }
                -
                101 
                -
                121 template <Backend B, Device D, class T>
                -
                122 void genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, D>& mat_a,
                -
                123  Matrix<T, D>& mat_b, Matrix<BaseType<T>, D>& eigenvalues,
                -
                124  Matrix<T, D>& eigenvectors) {
                -
                125  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
                -
                126  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
                -
                127  DLAF_ASSERT(matrix::local_matrix(eigenvalues), eigenvalues);
                -
                128  DLAF_ASSERT(matrix::equal_process_grid(eigenvectors, grid), eigenvectors, grid);
                -
                129  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                -
                130  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                -
                131  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                -
                132  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                -
                133  DLAF_ASSERT(matrix::square_size(eigenvectors), eigenvectors);
                -
                134  DLAF_ASSERT(matrix::square_blocksize(eigenvectors), eigenvectors);
                -
                135  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                -
                136  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                -
                137  DLAF_ASSERT(eigenvalues.size().rows() == eigenvectors.size().rows(), eigenvalues, eigenvectors);
                -
                138  DLAF_ASSERT(eigenvalues.blockSize().rows() == eigenvectors.blockSize().rows(), eigenvalues,
                -
                139  eigenvectors);
                -
                140  DLAF_ASSERT(eigenvectors.size() == mat_a.size(), eigenvectors, mat_a);
                -
                141  DLAF_ASSERT(eigenvectors.blockSize() == mat_a.blockSize(), eigenvectors, mat_a);
                -
                142 
                -
                143  internal::GenEigensolver<B, D, T>::call(grid, uplo, mat_a, mat_b, eigenvalues, eigenvectors);
                -
                144 }
                -
                145 
                -
                162 template <Backend B, Device D, class T>
                -
                163 EigensolverResult<T, D> genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, D>& mat_a,
                -
                164  Matrix<T, D>& mat_b) {
                -
                165  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
                -
                166  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
                -
                167  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                -
                168  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                -
                169  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                -
                170  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                -
                171  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                -
                172  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                -
                173 
                -
                174  const SizeType size = mat_a.size().rows();
                -
                175 
                -
                176  matrix::Matrix<BaseType<T>, D> eigenvalues(LocalElementSize(size, 1),
                -
                177  TileElementSize(mat_a.blockSize().rows(), 1));
                -
                178  matrix::Matrix<T, D> eigenvectors(GlobalElementSize(size, size), mat_a.blockSize(), grid);
                -
                179 
                -
                180  genEigensolver<B, D, T>(grid, uplo, mat_a, mat_b, eigenvalues, eigenvectors);
                +
                97  matrix::Matrix<BaseType<T>, D> eigenvalues(LocalElementSize(size, 1),
                +
                98  TileElementSize(mat_a.blockSize().rows(), 1));
                +
                99  matrix::Matrix<T, D> eigenvectors(LocalElementSize(size, size), mat_a.blockSize());
                +
                100 
                +
                101  genEigensolver<B, D, T>(uplo, mat_a, mat_b, eigenvalues, eigenvectors);
                +
                102 
                +
                103  return {std::move(eigenvalues), std::move(eigenvectors)};
                +
                104 }
                +
                105 
                +
                125 template <Backend B, Device D, class T>
                +
                126 void genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, D>& mat_a,
                +
                127  Matrix<T, D>& mat_b, Matrix<BaseType<T>, D>& eigenvalues,
                +
                128  Matrix<T, D>& eigenvectors) {
                +
                129  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
                +
                130  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
                +
                131  DLAF_ASSERT(matrix::local_matrix(eigenvalues), eigenvalues);
                +
                132  DLAF_ASSERT(matrix::equal_process_grid(eigenvectors, grid), eigenvectors, grid);
                +
                133  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                +
                134  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                +
                135  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                +
                136  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                +
                137  DLAF_ASSERT(matrix::square_size(eigenvectors), eigenvectors);
                +
                138  DLAF_ASSERT(matrix::square_blocksize(eigenvectors), eigenvectors);
                +
                139  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                +
                140  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                +
                141  DLAF_ASSERT(eigenvalues.size().rows() == eigenvectors.size().rows(), eigenvalues, eigenvectors);
                +
                142  DLAF_ASSERT(eigenvalues.blockSize().rows() == eigenvectors.blockSize().rows(), eigenvalues,
                +
                143  eigenvectors);
                +
                144  DLAF_ASSERT(eigenvectors.size() == mat_a.size(), eigenvectors, mat_a);
                +
                145  DLAF_ASSERT(eigenvectors.blockSize() == mat_a.blockSize(), eigenvectors, mat_a);
                +
                146  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
                +
                147  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
                +
                148  DLAF_ASSERT(matrix::single_tile_per_block(eigenvalues), eigenvalues);
                +
                149  DLAF_ASSERT(matrix::single_tile_per_block(eigenvectors), eigenvectors);
                +
                150 
                +
                151  internal::GenEigensolver<B, D, T>::call(grid, uplo, mat_a, mat_b, eigenvalues, eigenvectors);
                +
                152 }
                +
                153 
                +
                170 template <Backend B, Device D, class T>
                +
                171 EigensolverResult<T, D> genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, D>& mat_a,
                +
                172  Matrix<T, D>& mat_b) {
                +
                173  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
                +
                174  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
                +
                175  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                +
                176  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                +
                177  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                +
                178  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                +
                179  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                +
                180  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                181 
                -
                182  return {std::move(eigenvalues), std::move(eigenvectors)};
                -
                183 }
                -
                184 }
                +
                182  const SizeType size = mat_a.size().rows();
                +
                183 
                +
                184  matrix::Matrix<BaseType<T>, D> eigenvalues(LocalElementSize(size, 1),
                +
                185  TileElementSize(mat_a.blockSize().rows(), 1));
                +
                186  matrix::Matrix<T, D> eigenvectors(GlobalElementSize(size, size), mat_a.blockSize(), grid);
                +
                187 
                +
                188  genEigensolver<B, D, T>(grid, uplo, mat_a, mat_b, eigenvalues, eigenvectors);
                +
                189 
                +
                190  return {std::move(eigenvalues), std::move(eigenvectors)};
                +
                191 }
                +
                192 }
                dlaf::comm::CommunicatorGrid
                Definition: communicator_grid.h:42
                -
                dlaf::matrix::Matrix
                Definition: matrix.h:55
                +
                dlaf::matrix::Matrix
                Definition: matrix.h:56
                dlaf::eigensolver::genEigensolver
                void genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)
                Definition: gen_eigensolver.h:42
                diff --git a/master/gen__to__std_8h.html b/master/gen__to__std_8h.html index 99a2f4fe22..b7cd4771a9 100644 --- a/master/gen__to__std_8h.html +++ b/master/gen__to__std_8h.html @@ -137,6 +137,8 @@

                22 namespace dlaf {
                23 namespace eigensolver {
                24 
                -
                42 template <Backend backend, Device device, class T>
                -
                43 void genToStd(blas::Uplo uplo, Matrix<T, device>& mat_a, Matrix<T, device>& mat_b) {
                -
                44  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                -
                45  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                -
                46  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                -
                47  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                -
                48  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                -
                49  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                -
                50  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
                -
                51  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
                -
                52 
                -
                53  switch (uplo) {
                -
                54  case blas::Uplo::Lower:
                -
                55  internal::GenToStd<backend, device, T>::call_L(mat_a, mat_b);
                -
                56  break;
                -
                57  case blas::Uplo::Upper:
                -
                58  internal::GenToStd<backend, device, T>::call_U(mat_a, mat_b);
                +
                43 template <Backend backend, Device device, class T>
                +
                44 void genToStd(blas::Uplo uplo, Matrix<T, device>& mat_a, Matrix<T, device>& mat_b) {
                +
                45  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                +
                46  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                +
                47  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                +
                48  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                +
                49  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                +
                50  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                +
                51  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
                +
                52  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
                +
                53  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
                +
                54  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
                +
                55 
                +
                56  switch (uplo) {
                +
                57  case blas::Uplo::Lower:
                +
                58  internal::GenToStd<backend, device, T>::call_L(mat_a, mat_b);
                59  break;
                -
                60  case blas::Uplo::General:
                -
                61  DLAF_UNIMPLEMENTED(uplo);
                +
                60  case blas::Uplo::Upper:
                +
                61  internal::GenToStd<backend, device, T>::call_U(mat_a, mat_b);
                62  break;
                -
                63  }
                -
                64 }
                -
                65 
                -
                84 template <Backend backend, Device device, class T>
                -
                85 void genToStd(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, device>& mat_a,
                -
                86  Matrix<T, device>& mat_b) {
                -
                87  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                -
                88  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                -
                89  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                -
                90  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                -
                91  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                -
                92  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                -
                93  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
                -
                94  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
                -
                95 
                -
                96  switch (uplo) {
                -
                97  case blas::Uplo::Lower:
                -
                98  internal::GenToStd<backend, device, T>::call_L(grid, mat_a, mat_b);
                -
                99  break;
                -
                100  case blas::Uplo::Upper:
                -
                101  internal::GenToStd<backend, device, T>::call_U(grid, mat_a, mat_b);
                -
                102  break;
                -
                103  case blas::Uplo::General:
                -
                104  DLAF_UNIMPLEMENTED(uplo);
                +
                63  case blas::Uplo::General:
                +
                64  DLAF_UNIMPLEMENTED(uplo);
                +
                65  break;
                +
                66  }
                +
                67 }
                +
                68 
                +
                88 template <Backend backend, Device device, class T>
                +
                89 void genToStd(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix<T, device>& mat_a,
                +
                90  Matrix<T, device>& mat_b) {
                +
                91  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                +
                92  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                +
                93  DLAF_ASSERT(matrix::square_size(mat_b), mat_b);
                +
                94  DLAF_ASSERT(matrix::square_blocksize(mat_b), mat_b);
                +
                95  DLAF_ASSERT(mat_a.size() == mat_b.size(), mat_a, mat_b);
                +
                96  DLAF_ASSERT(mat_a.blockSize() == mat_b.blockSize(), mat_a, mat_b);
                +
                97  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
                +
                98  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
                +
                99  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
                +
                100  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
                +
                101 
                +
                102  switch (uplo) {
                +
                103  case blas::Uplo::Lower:
                +
                104  internal::GenToStd<backend, device, T>::call_L(grid, mat_a, mat_b);
                105  break;
                -
                106  }
                -
                107 }
                -
                108 
                -
                109 }
                -
                110 }
                +
                106  case blas::Uplo::Upper:
                +
                107  internal::GenToStd<backend, device, T>::call_U(grid, mat_a, mat_b);
                +
                108  break;
                +
                109  case blas::Uplo::General:
                +
                110  DLAF_UNIMPLEMENTED(uplo);
                +
                111  break;
                +
                112  }
                +
                113 }
                +
                114 
                +
                115 }
                +
                116 }
                dlaf::comm::CommunicatorGrid
                Definition: communicator_grid.h:42
                communicator_grid.h
                dlaf::eigensolver::eigensolver
                void eigensolver(blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)
                Definition: eigensolver.h:39
                -
                dlaf::eigensolver::genToStd
                void genToStd(blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)
                Definition: gen_to_std.h:43
                +
                dlaf::eigensolver::genToStd
                void genToStd(blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)
                Definition: gen_to_std.h:44
                diff --git a/master/hermitian_8h.html b/master/hermitian_8h.html index 57d7780bf6..25923c3f5b 100644 --- a/master/hermitian_8h.html +++ b/master/hermitian_8h.html @@ -162,6 +162,8 @@

                21 
                22 namespace dlaf::multiplication {
                23 
                -
                40 template <Backend B, Device D, class T>
                -
                41 void hermitian(blas::Side side, blas::Uplo uplo, const T alpha, Matrix<const T, D>& mat_a,
                -
                42  Matrix<const T, D>& mat_b, const T beta, Matrix<T, D>& mat_c) {
                -
                43  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                -
                44  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                -
                45  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
                -
                46  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
                -
                47  DLAF_ASSERT(matrix::local_matrix(mat_c), mat_c);
                -
                48 
                -
                49  if (side == blas::Side::Left) {
                -
                50  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_c, blas::Op::NoTrans, blas::Op::NoTrans), mat_a,
                -
                51  mat_b, mat_c);
                -
                52  switch (uplo) {
                -
                53  case blas::Uplo::Lower:
                -
                54  return internal::Hermitian<B, D, T>::call_LL(alpha, mat_a, mat_b, beta, mat_c);
                -
                55  break;
                -
                56  case blas::Uplo::Upper:
                -
                57  DLAF_UNIMPLEMENTED(uplo);
                -
                58  break;
                -
                59  case blas::Uplo::General:
                -
                60  DLAF_UNIMPLEMENTED(uplo);
                -
                61  break;
                -
                62  }
                -
                63  }
                -
                64  else {
                -
                65  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_c, blas::Op::NoTrans, blas::Op::NoTrans), mat_a,
                -
                66  mat_b, mat_c);
                -
                67  DLAF_UNIMPLEMENTED(side);
                -
                68  }
                -
                69 }
                -
                70 
                -
                88 template <Backend B, Device D, class T>
                -
                89 void hermitian(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, const T alpha,
                -
                90  Matrix<const T, D>& mat_a, Matrix<const T, D>& mat_b, const T beta, Matrix<T, D>& mat_c) {
                -
                91  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                -
                92  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                -
                93  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
                -
                94  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
                -
                95  DLAF_ASSERT(matrix::equal_process_grid(mat_c, grid), mat_c, grid);
                -
                96 
                -
                97  if (side == blas::Side::Left) {
                -
                98  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_c, blas::Op::NoTrans, blas::Op::NoTrans), mat_a,
                -
                99  mat_b, mat_c);
                -
                100  switch (uplo) {
                -
                101  case blas::Uplo::Lower:
                -
                102  return internal::Hermitian<B, D, T>::call_LL(grid, alpha, mat_a, mat_b, beta, mat_c);
                -
                103  break;
                -
                104  case blas::Uplo::Upper:
                -
                105  DLAF_UNIMPLEMENTED(uplo);
                -
                106  break;
                -
                107  case blas::Uplo::General:
                -
                108  DLAF_UNIMPLEMENTED(uplo);
                -
                109  break;
                -
                110  }
                -
                111  }
                -
                112  else {
                -
                113  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_c, blas::Op::NoTrans, blas::Op::NoTrans), mat_a,
                -
                114  mat_b, mat_c);
                -
                115  DLAF_UNIMPLEMENTED(side);
                -
                116  }
                -
                117 }
                -
                118 
                -
                119 }
                +
                41 template <Backend B, Device D, class T>
                +
                42 void hermitian(blas::Side side, blas::Uplo uplo, const T alpha, Matrix<const T, D>& mat_a,
                +
                43  Matrix<const T, D>& mat_b, const T beta, Matrix<T, D>& mat_c) {
                +
                44  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                +
                45  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                +
                46  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
                +
                47  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
                +
                48  DLAF_ASSERT(matrix::single_tile_per_block(mat_c), mat_c);
                +
                49  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
                +
                50  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
                +
                51  DLAF_ASSERT(matrix::local_matrix(mat_c), mat_c);
                +
                52 
                +
                53  if (side == blas::Side::Left) {
                +
                54  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_c, blas::Op::NoTrans, blas::Op::NoTrans), mat_a,
                +
                55  mat_b, mat_c);
                +
                56  switch (uplo) {
                +
                57  case blas::Uplo::Lower:
                +
                58  return internal::Hermitian<B, D, T>::call_LL(alpha, mat_a, mat_b, beta, mat_c);
                +
                59  break;
                +
                60  case blas::Uplo::Upper:
                +
                61  DLAF_UNIMPLEMENTED(uplo);
                +
                62  break;
                +
                63  case blas::Uplo::General:
                +
                64  DLAF_UNIMPLEMENTED(uplo);
                +
                65  break;
                +
                66  }
                +
                67  }
                +
                68  else {
                +
                69  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_c, blas::Op::NoTrans, blas::Op::NoTrans), mat_a,
                +
                70  mat_b, mat_c);
                +
                71  DLAF_UNIMPLEMENTED(side);
                +
                72  }
                +
                73 }
                +
                74 
                +
                93 template <Backend B, Device D, class T>
                +
                94 void hermitian(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, const T alpha,
                +
                95  Matrix<const T, D>& mat_a, Matrix<const T, D>& mat_b, const T beta, Matrix<T, D>& mat_c) {
                +
                96  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
                +
                97  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
                +
                98  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
                +
                99  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
                +
                100  DLAF_ASSERT(matrix::single_tile_per_block(mat_c), mat_c);
                +
                101  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
                +
                102  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
                +
                103  DLAF_ASSERT(matrix::equal_process_grid(mat_c, grid), mat_c, grid);
                +
                104 
                +
                105  if (side == blas::Side::Left) {
                +
                106  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_c, blas::Op::NoTrans, blas::Op::NoTrans), mat_a,
                +
                107  mat_b, mat_c);
                +
                108  switch (uplo) {
                +
                109  case blas::Uplo::Lower:
                +
                110  return internal::Hermitian<B, D, T>::call_LL(grid, alpha, mat_a, mat_b, beta, mat_c);
                +
                111  break;
                +
                112  case blas::Uplo::Upper:
                +
                113  DLAF_UNIMPLEMENTED(uplo);
                +
                114  break;
                +
                115  case blas::Uplo::General:
                +
                116  DLAF_UNIMPLEMENTED(uplo);
                +
                117  break;
                +
                118  }
                +
                119  }
                +
                120  else {
                +
                121  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_c, blas::Op::NoTrans, blas::Op::NoTrans), mat_a,
                +
                122  mat_b, mat_c);
                +
                123  DLAF_UNIMPLEMENTED(side);
                +
                124  }
                +
                125 }
                +
                126 
                +
                127 }
                dlaf::comm::CommunicatorGrid
                Definition: communicator_grid.h:42
                communicator_grid.h
                -
                dlaf::multiplication::hermitian
                void hermitian(blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)
                Definition: hermitian.h:41
                +
                dlaf::multiplication::hermitian
                void hermitian(blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)
                Definition: hermitian.h:42
                diff --git a/master/hierarchy.html b/master/hierarchy.html index e79c2658b6..d5e0443886 100644 --- a/master/hierarchy.html +++ b/master/hierarchy.html @@ -186,7 +186,6 @@

    - diff --git a/master/matrix_8h.html b/master/matrix_8h.html index a9abf17dd9..3365ad9a2f 100644 --- a/master/matrix_8h.html +++ b/master/matrix_8h.html @@ -77,6 +77,7 @@ #include <vector>
    #include <pika/execution.hpp>
    #include <dlaf/common/range2d.h>
    +#include <dlaf/common/vector.h>
    #include <dlaf/communication/communicator_grid.h>
    #include <dlaf/matrix/distribution.h>
    #include <dlaf/matrix/internal/tile_pipeline.h>
    diff --git a/master/matrix_8h_source.html b/master/matrix_8h_source.html index 25a8a83a15..f0482c09b8 100644 --- a/master/matrix_8h_source.html +++ b/master/matrix_8h_source.html @@ -88,254 +88,273 @@
    18 #include <pika/execution.hpp>
    19 
    20 #include <dlaf/common/range2d.h>
    -
    21 #include <dlaf/communication/communicator_grid.h>
    -
    22 #include <dlaf/matrix/distribution.h>
    -
    23 #include <dlaf/matrix/internal/tile_pipeline.h>
    -
    24 #include <dlaf/matrix/layout_info.h>
    -
    25 #include <dlaf/matrix/matrix_base.h>
    -
    26 #include <dlaf/matrix/tile.h>
    -
    27 #include <dlaf/types.h>
    -
    28 
    -
    29 namespace dlaf {
    -
    30 namespace matrix {
    -
    31 
    -
    32 namespace internal {
    -
    33 
    -
    37 template <class Func>
    -
    38 auto selectGeneric(Func&& f, common::IterableRange2D<SizeType, LocalTile_TAG> range) {
    -
    39  using RetT = decltype(f(LocalTileIndex{}));
    -
    40 
    -
    41  std::vector<RetT> tiles;
    -
    42  tiles.reserve(to_sizet(std::distance(range.begin(), range.end())));
    -
    43  std::transform(range.begin(), range.end(), std::back_inserter(tiles),
    -
    44  [&](auto idx) { return f(idx); });
    -
    45  return tiles;
    -
    46 }
    +
    21 #include <dlaf/common/vector.h>
    +
    22 #include <dlaf/communication/communicator_grid.h>
    +
    23 #include <dlaf/matrix/distribution.h>
    +
    24 #include <dlaf/matrix/internal/tile_pipeline.h>
    +
    25 #include <dlaf/matrix/layout_info.h>
    +
    26 #include <dlaf/matrix/matrix_base.h>
    +
    27 #include <dlaf/matrix/tile.h>
    +
    28 #include <dlaf/types.h>
    +
    29 
    +
    30 namespace dlaf {
    +
    31 namespace matrix {
    +
    32 
    +
    33 namespace internal {
    +
    34 
    +
    38 template <class Func>
    +
    39 auto selectGeneric(Func&& f, common::IterableRange2D<SizeType, LocalTile_TAG> range) {
    +
    40  using RetT = decltype(f(LocalTileIndex{}));
    +
    41 
    +
    42  std::vector<RetT> tiles;
    +
    43  tiles.reserve(to_sizet(std::distance(range.begin(), range.end())));
    +
    44  std::transform(range.begin(), range.end(), std::back_inserter(tiles),
    +
    45  [&](auto idx) { return f(idx); });
    +
    46  return tiles;
    47 }
    -
    48 
    -
    54 template <class T, Device D>
    -
    55 class Matrix : public Matrix<const T, D> {
    -
    56 public:
    -
    57  static constexpr Device device = D;
    -
    58 
    -
    59  using ElementType = T;
    -
    60  using TileType = Tile<ElementType, D>;
    -
    61  using ConstTileType = Tile<const ElementType, D>;
    -
    62  using TileDataType = internal::TileData<const ElementType, D>;
    -
    63  using ReadWriteSenderType = ReadWriteTileSender<T, D>;
    -
    64  friend Matrix<const ElementType, D>;
    -
    65 
    -
    70  Matrix(const LocalElementSize& size, const TileElementSize& block_size);
    -
    71 
    -
    77  Matrix(const GlobalElementSize& size, const TileElementSize& block_size,
    -
    78  const comm::CommunicatorGrid& comm);
    -
    79 
    -
    81  Matrix(Distribution distribution);
    -
    82 
    -
    90  Matrix(Distribution distribution, const LayoutInfo& layout) noexcept;
    -
    91 
    -
    99  Matrix(const LayoutInfo& layout, ElementType* ptr);
    -
    100 
    -
    110  Matrix(Distribution distribution, const LayoutInfo& layout, ElementType* ptr) noexcept;
    -
    111 
    -
    112  Matrix(const Matrix& rhs) = delete;
    -
    113  Matrix(Matrix&& rhs) = default;
    -
    114 
    -
    115  Matrix& operator=(const Matrix& rhs) = delete;
    -
    116  Matrix& operator=(Matrix&& rhs) = default;
    -
    117 
    -
    121  ReadWriteSenderType readwrite(const LocalTileIndex& index) noexcept {
    -
    122  return tile_managers_[tileLinearIndex(index)].readwrite();
    -
    123  }
    -
    124 
    -
    129  ReadWriteSenderType readwrite(const GlobalTileIndex& index) noexcept {
    -
    130  return readwrite(this->distribution().localTileIndex(index));
    -
    131  }
    -
    132 
    -
    133 private:
    -
    134  using typename Matrix<const T, D>::SubPipelineTag;
    -
    135  Matrix(Matrix& mat, const SubPipelineTag);
    -
    136 
    -
    137 public:
    -
    138  Matrix subPipeline() {
    -
    139  return Matrix(*this, SubPipelineTag{});
    -
    140  }
    -
    141 
    -
    142 protected:
    -
    143  using Matrix<const T, D>::tileLinearIndex;
    +
    48 }
    +
    49 
    +
    55 template <class T, Device D>
    +
    56 class Matrix : public Matrix<const T, D> {
    +
    57 public:
    +
    58  static constexpr Device device = D;
    +
    59 
    +
    60  using ElementType = T;
    +
    61  using TileType = Tile<ElementType, D>;
    +
    62  using ConstTileType = Tile<const ElementType, D>;
    +
    63  using TileDataType = internal::TileData<const ElementType, D>;
    +
    64  using ReadWriteSenderType = ReadWriteTileSender<T, D>;
    +
    65  friend Matrix<const ElementType, D>;
    +
    66 
    +
    71  Matrix(const LocalElementSize& size, const TileElementSize& block_size);
    +
    72 
    +
    78  Matrix(const GlobalElementSize& size, const TileElementSize& block_size,
    +
    79  const comm::CommunicatorGrid& comm);
    +
    80 
    +
    82  Matrix(Distribution distribution);
    +
    83 
    +
    91  Matrix(Distribution distribution, const LayoutInfo& layout) noexcept;
    +
    92 
    +
    100  Matrix(const LayoutInfo& layout, ElementType* ptr);
    +
    101 
    +
    111  Matrix(Distribution distribution, const LayoutInfo& layout, ElementType* ptr) noexcept;
    +
    112 
    +
    113  Matrix(const Matrix& rhs) = delete;
    +
    114  Matrix(Matrix&& rhs) = default;
    +
    115 
    +
    116  Matrix& operator=(const Matrix& rhs) = delete;
    +
    117  Matrix& operator=(Matrix&& rhs) = default;
    +
    118 
    +
    122  ReadWriteSenderType readwrite(const LocalTileIndex& index) noexcept {
    +
    123  return tile_managers_[tileLinearIndex(index)].readwrite();
    +
    124  }
    +
    125 
    +
    130  ReadWriteSenderType readwrite(const GlobalTileIndex& index) noexcept {
    +
    131  return readwrite(this->distribution().localTileIndex(index));
    +
    132  }
    +
    133 
    +
    134 public:
    +
    141  Matrix subPipeline() {
    +
    142  return Matrix(*this, SubPipelineTag{});
    +
    143  }
    144 
    -
    145 private:
    -
    146  using Matrix<const T, D>::setUpTiles;
    -
    147  using Matrix<const T, D>::tile_managers_;
    -
    148 };
    -
    149 
    -
    150 template <class T, Device D>
    -
    151 class Matrix<const T, D> : public internal::MatrixBase {
    -
    152 public:
    -
    153  static constexpr Device device = D;
    -
    154 
    -
    155  using ElementType = T;
    -
    156  using TileType = Tile<ElementType, D>;
    -
    157  using ConstTileType = Tile<const ElementType, D>;
    -
    158  using TileDataType = internal::TileData<ElementType, D>;
    -
    159  using ReadOnlySenderType = ReadOnlyTileSender<T, D>;
    -
    160  friend Matrix<ElementType, D>;
    -
    161 
    -
    162  Matrix(const LayoutInfo& layout, ElementType* ptr);
    -
    163 
    -
    164  Matrix(const LayoutInfo& layout, const ElementType* ptr)
    -
    165  : Matrix(layout, const_cast<ElementType*>(ptr)) {}
    -
    166 
    -
    167  Matrix(Distribution distribution, const LayoutInfo& layout, ElementType* ptr) noexcept;
    -
    168 
    -
    169  Matrix(Distribution distribution, const LayoutInfo& layout, const ElementType* ptr)
    -
    170  : Matrix(std::move(distribution), layout, const_cast<ElementType*>(ptr)) {}
    -
    171 
    -
    172  Matrix(const Matrix& rhs) = delete;
    -
    173  Matrix(Matrix&& rhs) = default;
    +
    154  Matrix retiledSubPipeline(const LocalTileSize& tiles_per_block) {
    +
    155  return Matrix(*this, tiles_per_block);
    +
    156  }
    +
    157 
    +
    158 protected:
    +
    159  using Matrix<const T, D>::tileLinearIndex;
    +
    160 
    +
    161 private:
    +
    162  using typename Matrix<const T, D>::SubPipelineTag;
    +
    163  Matrix(Matrix& mat, const SubPipelineTag);
    +
    164  Matrix(Matrix& mat, const LocalTileSize& tiles_per_block);
    +
    165 
    +
    166  using Matrix<const T, D>::setUpTiles;
    +
    167  using Matrix<const T, D>::tile_managers_;
    +
    168 };
    +
    169 
    +
    170 template <class T, Device D>
    +
    171 class Matrix<const T, D> : public internal::MatrixBase {
    +
    172 public:
    +
    173  static constexpr Device device = D;
    174 
    -
    175  Matrix& operator=(const Matrix& rhs) = delete;
    -
    176  Matrix& operator=(Matrix&& rhs) = default;
    -
    177 
    -
    181  ReadOnlySenderType read(const LocalTileIndex& index) noexcept {
    -
    182  return tile_managers_[tileLinearIndex(index)].read();
    -
    183  }
    -
    184 
    -
    189  ReadOnlySenderType read(const GlobalTileIndex& index) {
    -
    190  return read(distribution().localTileIndex(index));
    -
    191  }
    -
    192 
    -
    197  void waitLocalTiles() noexcept;
    -
    198 
    -
    199  Matrix subPipelineConst() {
    -
    200  return Matrix(*this, SubPipelineTag{});
    -
    201  }
    -
    202 
    -
    203 protected:
    -
    204  Matrix(Distribution distribution) : internal::MatrixBase{std::move(distribution)} {}
    -
    205 
    -
    206  struct SubPipelineTag {};
    -
    207  Matrix(Matrix& mat, const SubPipelineTag);
    -
    208 
    -
    209  void setUpTiles(const memory::MemoryView<ElementType, D>& mem, const LayoutInfo& layout) noexcept;
    -
    210  void setUpSubPipelines(Matrix<const T, D>&) noexcept;
    -
    211 
    -
    212  std::vector<internal::TilePipeline<T, D>> tile_managers_;
    -
    213 };
    -
    214 
    -
    215 // Note: the templates of the following helper functions are inverted w.r.t. the Matrix templates
    -
    216 // to allow the user to only specify the device and let the compiler deduce the type T.
    -
    217 
    -
    218 // Local versions
    -
    219 
    -
    229 template <Device D, class T>
    -
    230 Matrix<T, D> createMatrixFromColMajor(const LocalElementSize& size, const TileElementSize& block_size,
    -
    231  SizeType ld, T* ptr) {
    -
    232  return Matrix<T, D>(colMajorLayout(size, block_size, ld), ptr);
    -
    233 }
    -
    234 
    -
    242 template <Device D, class T>
    -
    243 Matrix<T, D> createMatrixFromTile(const LocalElementSize& size, const TileElementSize& block_size,
    -
    244  T* ptr) {
    -
    245  return Matrix<T, D>(tileLayout(size, block_size), ptr);
    -
    246 }
    -
    247 
    -
    259 template <Device D, class T>
    -
    260 Matrix<T, D> createMatrixFromTile(const LocalElementSize& size, const TileElementSize& block_size,
    -
    261  SizeType ld_tile, SizeType tiles_per_col, T* ptr) {
    -
    262  return Matrix<T, D>(tileLayout(size, block_size, ld_tile, tiles_per_col), ptr);
    -
    263 }
    +
    175  using ElementType = T;
    +
    176  using TileType = Tile<ElementType, D>;
    +
    177  using ConstTileType = Tile<const ElementType, D>;
    +
    178  using TileDataType = internal::TileData<ElementType, D>;
    +
    179  using ReadOnlySenderType = ReadOnlyTileSender<T, D>;
    +
    180  friend Matrix<ElementType, D>;
    +
    181 
    +
    182  Matrix(const LayoutInfo& layout, ElementType* ptr);
    +
    183 
    +
    184  Matrix(const LayoutInfo& layout, const ElementType* ptr)
    +
    185  : Matrix(layout, const_cast<ElementType*>(ptr)) {}
    +
    186 
    +
    187  Matrix(Distribution distribution, const LayoutInfo& layout, ElementType* ptr) noexcept;
    +
    188 
    +
    189  Matrix(Distribution distribution, const LayoutInfo& layout, const ElementType* ptr)
    +
    190  : Matrix(std::move(distribution), layout, const_cast<ElementType*>(ptr)) {}
    +
    191 
    +
    192  Matrix(const Matrix& rhs) = delete;
    +
    193  Matrix(Matrix&& rhs) = default;
    +
    194 
    +
    195  Matrix& operator=(const Matrix& rhs) = delete;
    +
    196  Matrix& operator=(Matrix&& rhs) = default;
    +
    197 
    +
    201  ReadOnlySenderType read(const LocalTileIndex& index) noexcept {
    +
    202  return tile_managers_[tileLinearIndex(index)].read();
    +
    203  }
    +
    204 
    +
    209  ReadOnlySenderType read(const GlobalTileIndex& index) {
    +
    210  return read(distribution().localTileIndex(index));
    +
    211  }
    +
    212 
    +
    217  void waitLocalTiles() noexcept;
    +
    218 
    +
    225  Matrix subPipelineConst() {
    +
    226  return Matrix(*this, SubPipelineTag{});
    +
    227  }
    +
    228 
    +
    238  Matrix retiledSubPipelineConst(const LocalTileSize& tiles_per_block) {
    +
    239  return Matrix(*this, tiles_per_block);
    +
    240  }
    +
    241 
    +
    246  void done(const LocalTileIndex& index) noexcept {
    +
    247  const auto i = tileLinearIndex(index);
    +
    248  tile_managers_[i].reset();
    +
    249  }
    +
    250 
    +
    255  void done(const GlobalTileIndex& index) noexcept {
    +
    256  done(distribution().localTileIndex(index));
    +
    257  }
    +
    258 
    +
    259 protected:
    +
    260  Matrix(Distribution distribution) : internal::MatrixBase{std::move(distribution)} {}
    +
    261  struct SubPipelineTag {};
    +
    262  Matrix(Matrix& mat, const SubPipelineTag);
    +
    263  Matrix(Matrix& mat, const LocalTileSize& tiles_per_block);
    264 
    -
    265 // Distributed versions
    -
    266 
    -
    278 template <Device D, class T>
    -
    279 Matrix<T, D> createMatrixFromColMajor(const GlobalElementSize& size, const TileElementSize& block_size,
    -
    280  SizeType ld, const comm::CommunicatorGrid& comm,
    -
    281  const comm::Index2D& source_rank_index, T* ptr) {
    -
    282  Distribution distribution(size, block_size, comm.size(), comm.rank(), source_rank_index);
    -
    283  auto layout = colMajorLayout(distribution.localSize(), block_size, ld);
    -
    284 
    -
    285  return Matrix<T, D>(std::move(distribution), layout, ptr);
    -
    286 }
    -
    287 
    -
    298 template <Device D, class T>
    -
    299 Matrix<T, D> createMatrixFromColMajor(const GlobalElementSize& size, const TileElementSize& block_size,
    -
    300  SizeType ld, const comm::CommunicatorGrid& comm, T* ptr) {
    -
    301  return createMatrixFromColMajor<D>(size, block_size, ld, comm, {0, 0}, ptr);
    -
    302 }
    -
    303 
    -
    313 template <Device D, class T>
    -
    314 Matrix<T, D> createMatrixFromTile(const GlobalElementSize& size, const TileElementSize& block_size,
    -
    315  const comm::CommunicatorGrid& comm,
    -
    316  const comm::Index2D& source_rank_index, T* ptr) {
    -
    317  Distribution distribution(size, block_size, comm.size(), comm.rank(), source_rank_index);
    -
    318  auto layout = tileLayout(distribution.localSize(), block_size);
    -
    319 
    -
    320  return Matrix<T, D>(std::move(distribution), layout, ptr);
    -
    321 }
    -
    322 
    -
    331 template <Device D, class T>
    -
    332 Matrix<T, D> createMatrixFromTile(const GlobalElementSize& size, const TileElementSize& block_size,
    -
    333  const comm::CommunicatorGrid& comm, T* ptr) {
    -
    334  return createMatrixFromTile<D>(size, block_size, comm, {0, 0}, ptr);
    -
    335 }
    -
    336 
    -
    350 template <Device D, class T>
    -
    351 Matrix<T, D> createMatrixFromTile(const GlobalElementSize& size, const TileElementSize& block_size,
    -
    352  SizeType ld_tile, SizeType tiles_per_col,
    -
    353  const comm::CommunicatorGrid& comm,
    -
    354  const comm::Index2D& source_rank_index, T* ptr) {
    -
    355  Distribution distribution(size, block_size, comm.size(), comm.rank(), source_rank_index);
    -
    356  auto layout = tileLayout(distribution.localSize(), block_size, ld_tile, tiles_per_col);
    -
    357 
    -
    358  return Matrix<T, D>(std::move(distribution), layout, ptr);
    +
    265  void setUpTiles(const memory::MemoryView<ElementType, D>& mem, const LayoutInfo& layout) noexcept;
    +
    266  void setUpSubPipelines(Matrix<const T, D>&) noexcept;
    +
    267  void setUpRetiledSubPipelines(Matrix<const T, D>&, const LocalTileSize& tiles_per_block) noexcept;
    +
    268 
    +
    269  std::vector<internal::TilePipeline<T, D>> tile_managers_;
    +
    270 };
    +
    271 
    +
    272 // Note: the templates of the following helper functions are inverted w.r.t. the Matrix templates
    +
    273 // to allow the user to only specify the device and let the compiler deduce the type T.
    +
    274 
    +
    275 // Local versions
    +
    276 
    +
    286 template <Device D, class T>
    +
    287 Matrix<T, D> createMatrixFromColMajor(const LocalElementSize& size, const TileElementSize& block_size,
    +
    288  SizeType ld, T* ptr) {
    +
    289  return Matrix<T, D>(colMajorLayout(size, block_size, ld), ptr);
    +
    290 }
    +
    291 
    +
    299 template <Device D, class T>
    +
    300 Matrix<T, D> createMatrixFromTile(const LocalElementSize& size, const TileElementSize& block_size,
    +
    301  T* ptr) {
    +
    302  return Matrix<T, D>(tileLayout(size, block_size), ptr);
    +
    303 }
    +
    304 
    +
    316 template <Device D, class T>
    +
    317 Matrix<T, D> createMatrixFromTile(const LocalElementSize& size, const TileElementSize& block_size,
    +
    318  SizeType ld_tile, SizeType tiles_per_col, T* ptr) {
    +
    319  return Matrix<T, D>(tileLayout(size, block_size, ld_tile, tiles_per_col), ptr);
    +
    320 }
    +
    321 
    +
    322 // Distributed versions
    +
    323 
    +
    335 template <Device D, class T>
    +
    336 Matrix<T, D> createMatrixFromColMajor(const GlobalElementSize& size, const TileElementSize& block_size,
    +
    337  SizeType ld, const comm::CommunicatorGrid& comm,
    +
    338  const comm::Index2D& source_rank_index, T* ptr) {
    +
    339  Distribution distribution(size, block_size, comm.size(), comm.rank(), source_rank_index);
    +
    340  auto layout = colMajorLayout(distribution.localSize(), block_size, ld);
    +
    341 
    +
    342  return Matrix<T, D>(std::move(distribution), layout, ptr);
    +
    343 }
    +
    344 
    +
    355 template <Device D, class T>
    +
    356 Matrix<T, D> createMatrixFromColMajor(const GlobalElementSize& size, const TileElementSize& block_size,
    +
    357  SizeType ld, const comm::CommunicatorGrid& comm, T* ptr) {
    +
    358  return createMatrixFromColMajor<D>(size, block_size, ld, comm, {0, 0}, ptr);
    359 }
    360 
    -
    373 template <Device D, class T>
    -
    374 Matrix<T, D> createMatrixFromTile(const GlobalElementSize& size, const TileElementSize& block_size,
    -
    375  SizeType ld_tile, SizeType tiles_per_col,
    -
    376  const comm::CommunicatorGrid& comm, T* ptr) {
    -
    377  return createMatrixFromTile<D>(size, block_size, ld_tile, tiles_per_col, comm, {0, 0}, ptr);
    +
    370 template <Device D, class T>
    +
    371 Matrix<T, D> createMatrixFromTile(const GlobalElementSize& size, const TileElementSize& block_size,
    +
    372  const comm::CommunicatorGrid& comm,
    +
    373  const comm::Index2D& source_rank_index, T* ptr) {
    +
    374  Distribution distribution(size, block_size, comm.size(), comm.rank(), source_rank_index);
    +
    375  auto layout = tileLayout(distribution.localSize(), block_size);
    +
    376 
    +
    377  return Matrix<T, D>(std::move(distribution), layout, ptr);
    378 }
    379 
    -
    383 template <class MatrixLike>
    -
    384 auto selectRead(MatrixLike& matrix, common::IterableRange2D<SizeType, LocalTile_TAG> range) {
    -
    385  return internal::selectGeneric([&](auto index) { return matrix.read(index); }, range);
    -
    386 }
    -
    387 
    -
    391 template <class MatrixLike>
    -
    392 auto select(MatrixLike& matrix, common::IterableRange2D<SizeType, LocalTile_TAG> range) {
    -
    393  return internal::selectGeneric([&](auto index) { return matrix.readwrite(index); }, range);
    -
    394 }
    -
    395 
    -
    396 // ETI
    -
    397 
    -
    398 #define DLAF_MATRIX_ETI(KWORD, DATATYPE, DEVICE) \
    -
    399  KWORD template class Matrix<DATATYPE, DEVICE>; \
    -
    400  KWORD template class Matrix<const DATATYPE, DEVICE>;
    -
    401 
    -
    402 DLAF_MATRIX_ETI(extern, float, Device::CPU)
    -
    403 DLAF_MATRIX_ETI(extern, double, Device::CPU)
    -
    404 DLAF_MATRIX_ETI(extern, std::complex<float>, Device::CPU)
    -
    405 DLAF_MATRIX_ETI(extern, std::complex<double>, Device::CPU)
    -
    406 
    -
    407 #if defined(DLAF_WITH_GPU)
    -
    408 DLAF_MATRIX_ETI(extern, float, Device::GPU)
    -
    409 DLAF_MATRIX_ETI(extern, double, Device::GPU)
    -
    410 DLAF_MATRIX_ETI(extern, std::complex<float>, Device::GPU)
    -
    411 DLAF_MATRIX_ETI(extern, std::complex<double>, Device::GPU)
    -
    412 #endif
    -
    413 }
    -
    414 #ifndef DLAF_DOXYGEN
    -
    415 // Note: Doxygen doesn't deal correctly with template specialized inheritance,
    -
    416 // and this line makes it run infinitely
    +
    388 template <Device D, class T>
    +
    389 Matrix<T, D> createMatrixFromTile(const GlobalElementSize& size, const TileElementSize& block_size,
    +
    390  const comm::CommunicatorGrid& comm, T* ptr) {
    +
    391  return createMatrixFromTile<D>(size, block_size, comm, {0, 0}, ptr);
    +
    392 }
    +
    393 
    +
    407 template <Device D, class T>
    +
    408 Matrix<T, D> createMatrixFromTile(const GlobalElementSize& size, const TileElementSize& block_size,
    +
    409  SizeType ld_tile, SizeType tiles_per_col,
    +
    410  const comm::CommunicatorGrid& comm,
    +
    411  const comm::Index2D& source_rank_index, T* ptr) {
    +
    412  Distribution distribution(size, block_size, comm.size(), comm.rank(), source_rank_index);
    +
    413  auto layout = tileLayout(distribution.localSize(), block_size, ld_tile, tiles_per_col);
    +
    414 
    +
    415  return Matrix<T, D>(std::move(distribution), layout, ptr);
    +
    416 }
    417 
    -
    419 using matrix::Matrix;
    -
    420 #endif
    -
    421 }
    -
    422 
    -
    423 #include <dlaf/matrix/matrix.tpp>
    -
    424 #include <dlaf/matrix/matrix_const.tpp>
    +
    430 template <Device D, class T>
    +
    431 Matrix<T, D> createMatrixFromTile(const GlobalElementSize& size, const TileElementSize& block_size,
    +
    432  SizeType ld_tile, SizeType tiles_per_col,
    +
    433  const comm::CommunicatorGrid& comm, T* ptr) {
    +
    434  return createMatrixFromTile<D>(size, block_size, ld_tile, tiles_per_col, comm, {0, 0}, ptr);
    +
    435 }
    +
    436 
    +
    440 template <class MatrixLike>
    +
    441 auto selectRead(MatrixLike& matrix, common::IterableRange2D<SizeType, LocalTile_TAG> range) {
    +
    442  return internal::selectGeneric([&](auto index) { return matrix.read(index); }, range);
    +
    443 }
    +
    444 
    +
    448 template <class MatrixLike>
    +
    449 auto select(MatrixLike& matrix, common::IterableRange2D<SizeType, LocalTile_TAG> range) {
    +
    450  return internal::selectGeneric([&](auto index) { return matrix.readwrite(index); }, range);
    +
    451 }
    +
    452 
    +
    453 // ETI
    +
    454 
    +
    455 #define DLAF_MATRIX_ETI(KWORD, DATATYPE, DEVICE) \
    +
    456  KWORD template class Matrix<DATATYPE, DEVICE>; \
    +
    457  KWORD template class Matrix<const DATATYPE, DEVICE>;
    +
    458 
    +
    459 DLAF_MATRIX_ETI(extern, float, Device::CPU)
    +
    460 DLAF_MATRIX_ETI(extern, double, Device::CPU)
    +
    461 DLAF_MATRIX_ETI(extern, std::complex<float>, Device::CPU)
    +
    462 DLAF_MATRIX_ETI(extern, std::complex<double>, Device::CPU)
    +
    463 
    +
    464 #if defined(DLAF_WITH_GPU)
    +
    465 DLAF_MATRIX_ETI(extern, float, Device::GPU)
    +
    466 DLAF_MATRIX_ETI(extern, double, Device::GPU)
    +
    467 DLAF_MATRIX_ETI(extern, std::complex<float>, Device::GPU)
    +
    468 DLAF_MATRIX_ETI(extern, std::complex<double>, Device::GPU)
    +
    469 #endif
    +
    470 }
    +
    471 #ifndef DLAF_DOXYGEN
    +
    472 // Note: Doxygen doesn't deal correctly with template specialized inheritance,
    +
    473 // and this line makes it run infinitely
    +
    474 
    +
    476 using matrix::Matrix;
    +
    477 #endif
    +
    478 }
    +
    479 
    +
    480 #include <dlaf/matrix/matrix.tpp>
    +
    481 #include <dlaf/matrix/matrix_const.tpp>
    dlaf::comm::CommunicatorGrid
    Definition: communicator_grid.h:42
    dlaf::comm::CommunicatorGrid::rank
    Index2D rank() const noexcept
    Definition: communicator_grid.h:63
    dlaf::comm::CommunicatorGrid::size
    Size2D size() const noexcept
    Return the number of rows in the grid.
    Definition: communicator_grid.h:68
    @@ -344,22 +363,26 @@
    dlaf::matrix::Distribution
    Definition: distribution.h:27
    dlaf::matrix::LayoutInfo
    Definition: layout_info.h:26
    -
    dlaf::matrix::Matrix< const T, D >
    Definition: matrix.h:151
    -
    dlaf::matrix::Matrix< const T, D >::read
    ReadOnlySenderType read(const GlobalTileIndex &index)
    Definition: matrix.h:189
    -
    dlaf::matrix::Matrix< const T, D >::read
    ReadOnlySenderType read(const LocalTileIndex &index) noexcept
    Definition: matrix.h:181
    +
    dlaf::matrix::Matrix< const T, D >
    Definition: matrix.h:171
    +
    dlaf::matrix::Matrix< const T, D >::read
    ReadOnlySenderType read(const GlobalTileIndex &index)
    Definition: matrix.h:209
    +
    dlaf::matrix::Matrix< const T, D >::retiledSubPipelineConst
    Matrix retiledSubPipelineConst(const LocalTileSize &tiles_per_block)
    Definition: matrix.h:238
    +
    dlaf::matrix::Matrix< const T, D >::done
    void done(const LocalTileIndex &index) noexcept
    Definition: matrix.h:246
    +
    dlaf::matrix::Matrix< const T, D >::read
    ReadOnlySenderType read(const LocalTileIndex &index) noexcept
    Definition: matrix.h:201
    -
    dlaf::matrix::Matrix
    Definition: matrix.h:55
    +
    dlaf::matrix::Matrix< const T, D >::done
    void done(const GlobalTileIndex &index) noexcept
    Definition: matrix.h:255
    +
    dlaf::matrix::Matrix
    Definition: matrix.h:56
    dlaf::matrix::Matrix::Matrix
    Matrix(Distribution distribution, const LayoutInfo &layout) noexcept
    dlaf::matrix::Matrix::Matrix
    Matrix(Distribution distribution, const LayoutInfo &layout, ElementType *ptr) noexcept
    dlaf::matrix::Matrix::Matrix
    Matrix(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm)
    +
    dlaf::matrix::Matrix::subPipeline
    Matrix subPipeline()
    Definition: matrix.h:141
    dlaf::matrix::Matrix::Matrix
    Matrix(Distribution distribution)
    Create a matrix distributed according to the distribution distribution.
    dlaf::matrix::Matrix::Matrix
    Matrix(const LocalElementSize &size, const TileElementSize &block_size)
    -
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
    Definition: matrix.h:121
    +
    dlaf::matrix::Matrix::retiledSubPipeline
    Matrix retiledSubPipeline(const LocalTileSize &tiles_per_block)
    Definition: matrix.h:154
    +
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
    Definition: matrix.h:122
    dlaf::matrix::Matrix::Matrix
    Matrix(const LayoutInfo &layout, ElementType *ptr)
    -
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const GlobalTileIndex &index) noexcept
    Definition: matrix.h:129
    +
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const GlobalTileIndex &index) noexcept
    Definition: matrix.h:130
    dlaf::matrix::Tile
    Definition: tile.h:329
    dlaf::matrix::internal::MatrixBase
    Definition: matrix_base.h:24
    -
    dlaf::matrix::internal::MatrixBase::tileLinearIndex
    std::size_t tileLinearIndex(const LocalTileIndex &index) const noexcept
    Definition: matrix_base.h:111
    dlaf::memory::MemoryView
    Definition: memory_view.h:32
    communicator_grid.h
    @@ -368,15 +391,16 @@
    dlaf::matrix::tileLayout
    LayoutInfo tileLayout(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col)
    Returns LayoutInfo for a local matrix which use the tile layout (Advanced interface).
    Definition: layout_info.h:135
    dlaf::matrix::colMajorLayout
    LayoutInfo colMajorLayout(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld)
    Returns LayoutInfo for a local column major matrix.
    Definition: layout_info.h:125
    -
    dlaf::matrix::selectRead
    auto selectRead(MatrixLike &matrix, common::IterableRange2D< SizeType, LocalTile_TAG > range)
    Definition: matrix.h:384
    -
    dlaf::matrix::createMatrixFromColMajor
    Matrix< T, D > createMatrixFromColMajor(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld, T *ptr)
    Definition: matrix.h:230
    -
    dlaf::matrix::select
    auto select(MatrixLike &matrix, common::IterableRange2D< SizeType, LocalTile_TAG > range)
    Definition: matrix.h:392
    -
    dlaf::matrix::createMatrixFromTile
    Matrix< T, D > createMatrixFromTile(const LocalElementSize &size, const TileElementSize &block_size, T *ptr)
    Definition: matrix.h:243
    -
    dlaf::matrix::internal::selectGeneric
    auto selectGeneric(Func &&f, common::IterableRange2D< SizeType, LocalTile_TAG > range)
    Definition: matrix.h:38
    +
    dlaf::matrix::selectRead
    auto selectRead(MatrixLike &matrix, common::IterableRange2D< SizeType, LocalTile_TAG > range)
    Definition: matrix.h:441
    +
    dlaf::matrix::createMatrixFromColMajor
    Matrix< T, D > createMatrixFromColMajor(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld, T *ptr)
    Definition: matrix.h:287
    +
    dlaf::matrix::select
    auto select(MatrixLike &matrix, common::IterableRange2D< SizeType, LocalTile_TAG > range)
    Definition: matrix.h:449
    +
    dlaf::matrix::createMatrixFromTile
    Matrix< T, D > createMatrixFromTile(const LocalElementSize &size, const TileElementSize &block_size, T *ptr)
    Definition: matrix.h:300
    +
    dlaf::matrix::internal::selectGeneric
    auto selectGeneric(Func &&f, common::IterableRange2D< SizeType, LocalTile_TAG > range)
    Definition: matrix.h:39
    matrix_base.h
    dlaf::to_sizet
    auto to_sizet(const T signed_value)
    Definition: types.h:265
    +
    diff --git a/master/multiplication_2general_8h.html b/master/multiplication_2general_8h.html index da0af44ffb..945a760519 100644 --- a/master/multiplication_2general_8h.html +++ b/master/multiplication_2general_8h.html @@ -187,6 +187,8 @@

    25 
    26 namespace dlaf::multiplication {
    27 
    -
    47 template <Backend B, Device D, class T>
    -
    48 void generalSubMatrix(const SizeType a, const SizeType b, const blas::Op opA, const blas::Op opB,
    -
    49  const T alpha, Matrix<const T, D>& mat_a, Matrix<const T, D>& mat_b, const T beta,
    -
    50  Matrix<T, D>& mat_c) {
    -
    51  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_a), mat_a);
    -
    52  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_b), mat_b);
    -
    53  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_c), mat_c);
    -
    54 
    -
    55  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
    -
    56  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
    -
    57  DLAF_ASSERT(matrix::local_matrix(mat_c), mat_c);
    -
    58 
    -
    59  // Note:
    -
    60  // This is an over-constraint, since the algorithm just cares about the sub-matrix size.
    -
    61  // It simplifies next check about [a,b) range validity, that otherwise would require it to be
    -
    62  // validated against every single sub-matrix in a, b and c that might have different element sizes.
    -
    63  //
    -
    64  // At the moment, we don't have this use-case, so let's keep it simple.
    -
    65  DLAF_ASSERT(dlaf::matrix::square_size(mat_a), mat_a);
    -
    66  DLAF_ASSERT(equal_size(mat_a, mat_b), mat_a, mat_b);
    -
    67  DLAF_ASSERT(equal_size(mat_a, mat_c), mat_a, mat_c);
    -
    68 
    -
    69  [[maybe_unused]] const SizeType m = mat_a.nrTiles().rows();
    -
    70  DLAF_ASSERT(a <= b, a, b);
    -
    71  DLAF_ASSERT(a >= 0 && a <= m, a, m);
    -
    72  DLAF_ASSERT(b >= 0 && b <= m, b, m);
    +
    48 template <Backend B, Device D, class T>
    +
    49 void generalSubMatrix(const SizeType a, const SizeType b, const blas::Op opA, const blas::Op opB,
    +
    50  const T alpha, Matrix<const T, D>& mat_a, Matrix<const T, D>& mat_b, const T beta,
    +
    51  Matrix<T, D>& mat_c) {
    +
    52  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_a), mat_a);
    +
    53  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_b), mat_b);
    +
    54  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_c), mat_c);
    +
    55 
    +
    56  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
    +
    57  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
    +
    58  DLAF_ASSERT(matrix::single_tile_per_block(mat_c), mat_c);
    +
    59 
    +
    60  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
    +
    61  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
    +
    62  DLAF_ASSERT(matrix::local_matrix(mat_c), mat_c);
    +
    63 
    +
    64  // Note:
    +
    65  // This is an over-constraint, since the algorithm just cares about the sub-matrix size.
    +
    66  // It simplifies next check about [a,b) range validity, that otherwise would require it to be
    +
    67  // validated against every single sub-matrix in a, b and c that might have different element sizes.
    +
    68  //
    +
    69  // At the moment, we don't have this use-case, so let's keep it simple.
    +
    70  DLAF_ASSERT(dlaf::matrix::square_size(mat_a), mat_a);
    +
    71  DLAF_ASSERT(equal_size(mat_a, mat_b), mat_a, mat_b);
    +
    72  DLAF_ASSERT(equal_size(mat_a, mat_c), mat_a, mat_c);
    73 
    -
    74  using namespace blas;
    -
    75 
    -
    76  if (opA == Op::NoTrans && opB == Op::NoTrans)
    -
    77  internal::GeneralSub<B, D, T>::callNN(a, b, opA, opB, alpha, mat_a, mat_b, beta, mat_c);
    -
    78  else
    -
    79  DLAF_UNIMPLEMENTED(opA, opB);
    -
    80 }
    -
    81 
    -
    97 template <Backend B, Device D, class T>
    -
    98 void generalSubMatrix([[maybe_unused]] comm::CommunicatorGrid grid,
    -
    99  common::Pipeline<comm::Communicator>& row_task_chain,
    -
    100  common::Pipeline<comm::Communicator>& col_task_chain, const SizeType a,
    -
    101  const SizeType b, const T alpha, Matrix<const T, D>& mat_a,
    -
    102  Matrix<const T, D>& mat_b, const T beta, Matrix<T, D>& mat_c) {
    -
    103  DLAF_ASSERT(equal_process_grid(mat_a, grid), mat_a, grid);
    -
    104  DLAF_ASSERT(equal_process_grid(mat_b, grid), mat_a, grid);
    -
    105  DLAF_ASSERT(equal_process_grid(mat_c, grid), mat_a, grid);
    -
    106 
    -
    107  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_a), mat_a);
    -
    108  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_b), mat_b);
    -
    109  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_c), mat_c);
    -
    110 
    -
    111  // Note:
    -
    112  // This is an over-constraint, since the algorithm just cares about the sub-matrix size (and its
    -
    113  // distribution).
    -
    114  // It simplifies next check about [a,b) range validity, that otherwise would require it to be
    -
    115  // validated against every single sub-matrix in a, b and c that might have different element sizes.
    -
    116  //
    -
    117  // At the moment, we don't have this use-case, so let's keep it simple.
    -
    118  DLAF_ASSERT(dlaf::matrix::square_size(mat_a), mat_a);
    -
    119  DLAF_ASSERT(equal_size(mat_a, mat_b), mat_a, mat_b);
    -
    120  DLAF_ASSERT(equal_size(mat_a, mat_c), mat_a, mat_c);
    -
    121 
    -
    122  [[maybe_unused]] const SizeType m = mat_a.nrTiles().rows();
    -
    123  DLAF_ASSERT(a <= b, a, b);
    -
    124  DLAF_ASSERT(a >= 0 && a <= m, a, m);
    -
    125  DLAF_ASSERT(b >= 0 && b <= m, b, m);
    -
    126 
    -
    127  internal::GeneralSub<B, D, T>::callNN(row_task_chain, col_task_chain, a, b, alpha, mat_a, mat_b, beta,
    -
    128  mat_c);
    -
    129 }
    -
    130 
    -
    131 template <Backend B, Device D, class T>
    -
    132 void generalSubMatrix(comm::CommunicatorGrid grid, const SizeType a, const SizeType b, const T alpha,
    -
    133  Matrix<const T, D>& mat_a, Matrix<const T, D>& mat_b, const T beta,
    -
    134  Matrix<T, D>& mat_c) {
    -
    135  common::Pipeline<comm::Communicator> row_task_chain(grid.rowCommunicator().clone());
    -
    136  common::Pipeline<comm::Communicator> col_task_chain(grid.colCommunicator().clone());
    -
    137  generalSubMatrix<B, D, T>(grid, row_task_chain, col_task_chain, a, b, alpha, mat_a, mat_b, beta,
    -
    138  mat_c);
    +
    74  [[maybe_unused]] const SizeType m = mat_a.nrTiles().rows();
    +
    75  DLAF_ASSERT(a <= b, a, b);
    +
    76  DLAF_ASSERT(a >= 0 && a <= m, a, m);
    +
    77  DLAF_ASSERT(b >= 0 && b <= m, b, m);
    +
    78 
    +
    79  using namespace blas;
    +
    80 
    +
    81  if (opA == Op::NoTrans && opB == Op::NoTrans)
    +
    82  internal::GeneralSub<B, D, T>::callNN(a, b, opA, opB, alpha, mat_a, mat_b, beta, mat_c);
    +
    83  else
    +
    84  DLAF_UNIMPLEMENTED(opA, opB);
    +
    85 }
    +
    86 
    +
    103 template <Backend B, Device D, class T>
    +
    104 void generalSubMatrix([[maybe_unused]] comm::CommunicatorGrid grid,
    +
    105  common::Pipeline<comm::Communicator>& row_task_chain,
    +
    106  common::Pipeline<comm::Communicator>& col_task_chain, const SizeType a,
    +
    107  const SizeType b, const T alpha, Matrix<const T, D>& mat_a,
    +
    108  Matrix<const T, D>& mat_b, const T beta, Matrix<T, D>& mat_c) {
    +
    109  DLAF_ASSERT(equal_process_grid(mat_a, grid), mat_a, grid);
    +
    110  DLAF_ASSERT(equal_process_grid(mat_b, grid), mat_a, grid);
    +
    111  DLAF_ASSERT(equal_process_grid(mat_c, grid), mat_a, grid);
    +
    112 
    +
    113  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_a), mat_a);
    +
    114  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_b), mat_b);
    +
    115  DLAF_ASSERT(dlaf::matrix::square_blocksize(mat_c), mat_c);
    +
    116 
    +
    117  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
    +
    118  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
    +
    119  DLAF_ASSERT(matrix::single_tile_per_block(mat_c), mat_c);
    +
    120 
    +
    121  // Note:
    +
    122  // This is an over-constraint, since the algorithm just cares about the sub-matrix size (and its
    +
    123  // distribution).
    +
    124  // It simplifies next check about [a,b) range validity, that otherwise would require it to be
    +
    125  // validated against every single sub-matrix in a, b and c that might have different element sizes.
    +
    126  //
    +
    127  // At the moment, we don't have this use-case, so let's keep it simple.
    +
    128  DLAF_ASSERT(dlaf::matrix::square_size(mat_a), mat_a);
    +
    129  DLAF_ASSERT(equal_size(mat_a, mat_b), mat_a, mat_b);
    +
    130  DLAF_ASSERT(equal_size(mat_a, mat_c), mat_a, mat_c);
    +
    131 
    +
    132  [[maybe_unused]] const SizeType m = mat_a.nrTiles().rows();
    +
    133  DLAF_ASSERT(a <= b, a, b);
    +
    134  DLAF_ASSERT(a >= 0 && a <= m, a, m);
    +
    135  DLAF_ASSERT(b >= 0 && b <= m, b, m);
    +
    136 
    +
    137  internal::GeneralSub<B, D, T>::callNN(row_task_chain, col_task_chain, a, b, alpha, mat_a, mat_b, beta,
    +
    138  mat_c);
    139 }
    140 
    -
    141 }
    +
    141 template <Backend B, Device D, class T>
    +
    142 void generalSubMatrix(comm::CommunicatorGrid grid, const SizeType a, const SizeType b, const T alpha,
    +
    143  Matrix<const T, D>& mat_a, Matrix<const T, D>& mat_b, const T beta,
    +
    144  Matrix<T, D>& mat_c) {
    +
    145  common::Pipeline<comm::Communicator> row_task_chain(grid.rowCommunicator().clone());
    +
    146  common::Pipeline<comm::Communicator> col_task_chain(grid.colCommunicator().clone());
    +
    147  generalSubMatrix<B, D, T>(grid, row_task_chain, col_task_chain, a, b, alpha, mat_a, mat_b, beta,
    +
    148  mat_c);
    +
    149 }
    +
    150 
    +
    151 }
    dlaf::comm::CommunicatorGrid
    Definition: communicator_grid.h:42
    dlaf::comm::CommunicatorGrid::rowCommunicator
    Communicator & rowCommunicator() noexcept
    Return a Communicator grouping all ranks in the row (that includes the current process).
    Definition: communicator_grid.h:78
    dlaf::comm::CommunicatorGrid::colCommunicator
    Communicator & colCommunicator() noexcept
    Return a Communicator grouping all ranks in the column (that includes the current process).
    Definition: communicator_grid.h:83
    @@ -184,7 +192,7 @@ -
    dlaf::multiplication::generalSubMatrix
    void generalSubMatrix(const SizeType a, const SizeType b, const blas::Op opA, const blas::Op opB, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)
    Definition: general.h:48
    +
    dlaf::multiplication::generalSubMatrix
    void generalSubMatrix(const SizeType a, const SizeType b, const blas::Op opA, const blas::Op opB, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)
    Definition: general.h:49
    util_matrix.h
    diff --git a/master/multiplication_2triangular_8h.html b/master/multiplication_2triangular_8h.html index a4b7425141..7198174257 100644 --- a/master/multiplication_2triangular_8h.html +++ b/master/multiplication_2triangular_8h.html @@ -163,6 +163,8 @@

    22 namespace dlaf {
    23 namespace multiplication {
    24 
    -
    42 template <Backend backend, Device device, class T>
    -
    43 void triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha,
    -
    44  Matrix<const T, device>& mat_a, Matrix<T, device>& mat_b) {
    -
    45  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    -
    46  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    -
    47  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
    -
    48  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
    -
    49 
    -
    50  if (side == blas::Side::Left) {
    -
    51  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_b, op, blas::Op::NoTrans), mat_a, mat_b, op);
    +
    43 template <Backend backend, Device device, class T>
    +
    44 void triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha,
    +
    45  Matrix<const T, device>& mat_a, Matrix<T, device>& mat_b) {
    +
    46  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    +
    47  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    +
    48  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
    +
    49  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
    +
    50  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
    +
    51  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
    52 
    -
    53  if (uplo == blas::Uplo::Lower) {
    -
    54  if (op == blas::Op::NoTrans) {
    -
    55  internal::Triangular<backend, device, T>::call_LLN(diag, alpha, mat_a, mat_b);
    -
    56  }
    -
    57  else {
    -
    58  internal::Triangular<backend, device, T>::call_LLT(op, diag, alpha, mat_a, mat_b);
    +
    53  if (side == blas::Side::Left) {
    +
    54  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_b, op, blas::Op::NoTrans), mat_a, mat_b, op);
    +
    55 
    +
    56  if (uplo == blas::Uplo::Lower) {
    +
    57  if (op == blas::Op::NoTrans) {
    +
    58  internal::Triangular<backend, device, T>::call_LLN(diag, alpha, mat_a, mat_b);
    59  }
    -
    60  }
    -
    61  else {
    -
    62  if (op == blas::Op::NoTrans) {
    -
    63  internal::Triangular<backend, device, T>::call_LUN(diag, alpha, mat_a, mat_b);
    -
    64  }
    -
    65  else {
    -
    66  internal::Triangular<backend, device, T>::call_LUT(op, diag, alpha, mat_a, mat_b);
    +
    60  else {
    +
    61  internal::Triangular<backend, device, T>::call_LLT(op, diag, alpha, mat_a, mat_b);
    +
    62  }
    +
    63  }
    +
    64  else {
    +
    65  if (op == blas::Op::NoTrans) {
    +
    66  internal::Triangular<backend, device, T>::call_LUN(diag, alpha, mat_a, mat_b);
    67  }
    -
    68  }
    -
    69  }
    -
    70  else {
    -
    71  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_b, blas::Op::NoTrans, op), mat_a, mat_b, op);
    -
    72 
    -
    73  if (uplo == blas::Uplo::Lower) {
    -
    74  if (op == blas::Op::NoTrans) {
    -
    75  internal::Triangular<backend, device, T>::call_RLN(diag, alpha, mat_a, mat_b);
    -
    76  }
    -
    77  else {
    -
    78  internal::Triangular<backend, device, T>::call_RLT(op, diag, alpha, mat_a, mat_b);
    +
    68  else {
    +
    69  internal::Triangular<backend, device, T>::call_LUT(op, diag, alpha, mat_a, mat_b);
    +
    70  }
    +
    71  }
    +
    72  }
    +
    73  else {
    +
    74  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_b, blas::Op::NoTrans, op), mat_a, mat_b, op);
    +
    75 
    +
    76  if (uplo == blas::Uplo::Lower) {
    +
    77  if (op == blas::Op::NoTrans) {
    +
    78  internal::Triangular<backend, device, T>::call_RLN(diag, alpha, mat_a, mat_b);
    79  }
    -
    80  }
    -
    81  else {
    -
    82  if (op == blas::Op::NoTrans) {
    -
    83  internal::Triangular<backend, device, T>::call_RUN(diag, alpha, mat_a, mat_b);
    -
    84  }
    -
    85  else {
    -
    86  internal::Triangular<backend, device, T>::call_RUT(op, diag, alpha, mat_a, mat_b);
    +
    80  else {
    +
    81  internal::Triangular<backend, device, T>::call_RLT(op, diag, alpha, mat_a, mat_b);
    +
    82  }
    +
    83  }
    +
    84  else {
    +
    85  if (op == blas::Op::NoTrans) {
    +
    86  internal::Triangular<backend, device, T>::call_RUN(diag, alpha, mat_a, mat_b);
    87  }
    -
    88  }
    -
    89  }
    -
    90 }
    -
    91 
    -
    109 template <Backend backend, Device device, class T>
    -
    110 void triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op,
    -
    111  blas::Diag diag, T alpha, Matrix<const T, device>& mat_a, Matrix<T, device>& mat_b) {
    -
    112  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    -
    113  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    -
    114  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
    -
    115  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
    -
    116 
    -
    117  if (side == blas::Side::Left) {
    -
    118  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_b, op, blas::Op::NoTrans), mat_a, mat_b, op);
    -
    119 
    -
    120  if (uplo == blas::Uplo::Lower) {
    -
    121  if (op == blas::Op::NoTrans) {
    -
    122  internal::Triangular<backend, device, T>::call_LLN(grid, diag, alpha, mat_a, mat_b);
    -
    123  }
    -
    124  else {
    -
    125  // Left Lower Trans/ConjTrans
    -
    126  DLAF_UNIMPLEMENTED(side, uplo, op, diag);
    -
    127  }
    -
    128  }
    -
    129  else {
    -
    130  if (op == blas::Op::NoTrans) {
    -
    131  internal::Triangular<backend, device, T>::call_LUN(grid, diag, alpha, mat_a, mat_b);
    -
    132  }
    -
    133  else {
    -
    134  // Left Upper Trans/ConjTrans
    -
    135  DLAF_UNIMPLEMENTED(side, uplo, op, diag);
    -
    136  }
    -
    137  }
    -
    138  }
    -
    139  else {
    -
    140  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_b, blas::Op::NoTrans, op), mat_a, mat_b, op);
    -
    141 
    -
    142  if (uplo == blas::Uplo::Lower) {
    -
    143  if (op == blas::Op::NoTrans) {
    -
    144  internal::Triangular<backend, device, T>::call_RLN(grid, diag, alpha, mat_a, mat_b);
    -
    145  }
    -
    146  else {
    -
    147  // Right Lower Trans/ConjTrans
    -
    148  DLAF_UNIMPLEMENTED(side, uplo, op, diag);
    -
    149  }
    -
    150  }
    -
    151  else {
    -
    152  if (op == blas::Op::NoTrans) {
    -
    153  internal::Triangular<backend, device, T>::call_RUN(grid, diag, alpha, mat_a, mat_b);
    -
    154  }
    -
    155  else {
    -
    156  // Right Upper Trans/ConjTrans
    -
    157  DLAF_UNIMPLEMENTED(side, uplo, op, diag);
    -
    158  }
    -
    159  }
    -
    160  }
    -
    161 }
    -
    162 
    -
    163 }
    -
    164 }
    +
    88  else {
    +
    89  internal::Triangular<backend, device, T>::call_RUT(op, diag, alpha, mat_a, mat_b);
    +
    90  }
    +
    91  }
    +
    92  }
    +
    93 }
    +
    94 
    +
    113 template <Backend backend, Device device, class T>
    +
    114 void triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op,
    +
    115  blas::Diag diag, T alpha, Matrix<const T, device>& mat_a, Matrix<T, device>& mat_b) {
    +
    116  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    +
    117  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    +
    118  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
    +
    119  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
    +
    120  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
    +
    121  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
    +
    122 
    +
    123  if (side == blas::Side::Left) {
    +
    124  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_b, op, blas::Op::NoTrans), mat_a, mat_b, op);
    +
    125 
    +
    126  if (uplo == blas::Uplo::Lower) {
    +
    127  if (op == blas::Op::NoTrans) {
    +
    128  internal::Triangular<backend, device, T>::call_LLN(grid, diag, alpha, mat_a, mat_b);
    +
    129  }
    +
    130  else {
    +
    131  // Left Lower Trans/ConjTrans
    +
    132  DLAF_UNIMPLEMENTED(side, uplo, op, diag);
    +
    133  }
    +
    134  }
    +
    135  else {
    +
    136  if (op == blas::Op::NoTrans) {
    +
    137  internal::Triangular<backend, device, T>::call_LUN(grid, diag, alpha, mat_a, mat_b);
    +
    138  }
    +
    139  else {
    +
    140  // Left Upper Trans/ConjTrans
    +
    141  DLAF_UNIMPLEMENTED(side, uplo, op, diag);
    +
    142  }
    +
    143  }
    +
    144  }
    +
    145  else {
    +
    146  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_b, blas::Op::NoTrans, op), mat_a, mat_b, op);
    +
    147 
    +
    148  if (uplo == blas::Uplo::Lower) {
    +
    149  if (op == blas::Op::NoTrans) {
    +
    150  internal::Triangular<backend, device, T>::call_RLN(grid, diag, alpha, mat_a, mat_b);
    +
    151  }
    +
    152  else {
    +
    153  // Right Lower Trans/ConjTrans
    +
    154  DLAF_UNIMPLEMENTED(side, uplo, op, diag);
    +
    155  }
    +
    156  }
    +
    157  else {
    +
    158  if (op == blas::Op::NoTrans) {
    +
    159  internal::Triangular<backend, device, T>::call_RUN(grid, diag, alpha, mat_a, mat_b);
    +
    160  }
    +
    161  else {
    +
    162  // Right Upper Trans/ConjTrans
    +
    163  DLAF_UNIMPLEMENTED(side, uplo, op, diag);
    +
    164  }
    +
    165  }
    +
    166  }
    +
    167 }
    +
    168 
    +
    169 }
    +
    170 }
    dlaf::comm::CommunicatorGrid
    Definition: communicator_grid.h:42
    communicator_grid.h
    -
    dlaf::multiplication::triangular
    void triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)
    Definition: triangular.h:43
    +
    dlaf::multiplication::triangular
    void triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)
    Definition: triangular.h:44
    util_matrix.h
    diff --git a/master/norm_8h.html b/master/norm_8h.html index c7f8e6f77d..a4219880e8 100644 --- a/master/norm_8h.html +++ b/master/norm_8h.html @@ -149,7 +149,9 @@

    Note
    not yet implemented
    Precondition
    A.blockSize().rows() == A.blockSize().cols(),
    -A is distributed according to grid,
    +A is distributed according to grid, +
    +A has equal tile and block sizes,
    Returns
    the norm norm_type of the Matrix A or 0 if A.size().isEmpty() (see LAPACK doc for additional info).
    diff --git a/master/norm_8h_source.html b/master/norm_8h_source.html index df04f2b0cb..e930d0e8a8 100644 --- a/master/norm_8h_source.html +++ b/master/norm_8h_source.html @@ -93,49 +93,51 @@
    23 
    24 namespace dlaf::auxiliary {
    25 
    -
    42 template <Backend backend, Device device, class T>
    -
    43 dlaf::BaseType<T> norm(comm::CommunicatorGrid grid, comm::Index2D rank, lapack::Norm norm_type,
    -
    44  blas::Uplo uplo, Matrix<const T, device>& A) {
    -
    45  using dlaf::matrix::equal_process_grid;
    -
    46 
    -
    47  DLAF_ASSERT(equal_process_grid(A, grid), A, grid);
    +
    43 template <Backend backend, Device device, class T>
    +
    44 dlaf::BaseType<T> norm(comm::CommunicatorGrid grid, comm::Index2D rank, lapack::Norm norm_type,
    +
    45  blas::Uplo uplo, Matrix<const T, device>& A) {
    +
    46  using dlaf::matrix::equal_process_grid;
    +
    47  using dlaf::matrix::single_tile_per_block;
    48 
    -
    49  // LAPACK documentation specify that if any dimension is 0, the result is 0
    -
    50  if (A.size().isEmpty())
    -
    51  return {0};
    -
    52 
    -
    53  switch (norm_type) {
    -
    54  case lapack::Norm::One:
    -
    55  case lapack::Norm::Two:
    -
    56  case lapack::Norm::Inf:
    -
    57  case lapack::Norm::Fro:
    -
    58  DLAF_UNIMPLEMENTED(norm_type);
    -
    59  return {};
    -
    60  case lapack::Norm::Max:
    -
    61  switch (uplo) {
    -
    62  case blas::Uplo::Lower:
    -
    63  return internal::Norm<backend, device, T>::max_L(grid, rank, A);
    -
    64  case blas::Uplo::Upper:
    -
    65  DLAF_UNIMPLEMENTED(norm_type, uplo);
    -
    66  return {};
    -
    67  case blas::Uplo::General:
    -
    68  return internal::Norm<backend, device, T>::max_G(grid, rank, A);
    -
    69  default:
    -
    70  return {};
    -
    71  }
    -
    72  default:
    -
    73  return {};
    -
    74  }
    -
    75 }
    -
    76 
    -
    77 }
    +
    49  DLAF_ASSERT(equal_process_grid(A, grid), A, grid);
    +
    50  DLAF_ASSERT(single_tile_per_block(A), A);
    +
    51 
    +
    52  // LAPACK documentation specify that if any dimension is 0, the result is 0
    +
    53  if (A.size().isEmpty())
    +
    54  return {0};
    +
    55 
    +
    56  switch (norm_type) {
    +
    57  case lapack::Norm::One:
    +
    58  case lapack::Norm::Two:
    +
    59  case lapack::Norm::Inf:
    +
    60  case lapack::Norm::Fro:
    +
    61  DLAF_UNIMPLEMENTED(norm_type);
    +
    62  return {};
    +
    63  case lapack::Norm::Max:
    +
    64  switch (uplo) {
    +
    65  case blas::Uplo::Lower:
    +
    66  return internal::Norm<backend, device, T>::max_L(grid, rank, A);
    +
    67  case blas::Uplo::Upper:
    +
    68  DLAF_UNIMPLEMENTED(norm_type, uplo);
    +
    69  return {};
    +
    70  case blas::Uplo::General:
    +
    71  return internal::Norm<backend, device, T>::max_G(grid, rank, A);
    +
    72  default:
    +
    73  return {};
    +
    74  }
    +
    75  default:
    +
    76  return {};
    +
    77  }
    +
    78 }
    +
    79 
    +
    80 }
    enum_output.h
    dlaf::comm::CommunicatorGrid
    Definition: communicator_grid.h:42
    communicator_grid.h
    -
    dlaf::auxiliary::norm
    dlaf::BaseType< T > norm(comm::CommunicatorGrid grid, comm::Index2D rank, lapack::Norm norm_type, blas::Uplo uplo, Matrix< const T, device > &A)
    Definition: norm.h:43
    +
    dlaf::auxiliary::norm
    dlaf::BaseType< T > norm(comm::CommunicatorGrid grid, comm::Index2D rank, lapack::Norm norm_type, blas::Uplo uplo, Matrix< const T, device > &A)
    Definition: norm.h:44
    util_matrix.h
    diff --git a/master/panel_8h_source.html b/master/panel_8h_source.html index 746915844c..090adbd2d2 100644 --- a/master/panel_8h_source.html +++ b/master/panel_8h_source.html @@ -552,7 +552,7 @@
    dlaf::common::internal::basic_coords::get
    IndexT get() const noexcept
    Return a copy of the row or the col index as specified by rc.
    Definition: index2d.h:86
    dlaf::common::internal::basic_coords::transpose
    void transpose() noexcept
    Swaps row and column index/size.
    Definition: index2d.h:106
    dlaf::matrix::Distribution
    Definition: distribution.h:27
    -
    dlaf::matrix::Matrix
    Definition: matrix.h:55
    +
    dlaf::matrix::Matrix
    Definition: matrix.h:56
    dlaf::matrix::Tile< const T, D >
    Definition: tile.h:201
    dlaf::matrix::Tile
    Definition: tile.h:329
    diff --git a/master/permutations_2general_2impl_8h_source.html b/master/permutations_2general_2impl_8h_source.html index b9c31c3b12..a350e08801 100644 --- a/master/permutations_2general_2impl_8h_source.html +++ b/master/permutations_2general_2impl_8h_source.html @@ -655,8 +655,8 @@
    dlaf::matrix::Distribution::globalElementFromGlobalTileAndTileElement
    SizeType globalElementFromGlobalTileAndTileElement(SizeType global_tile, SizeType tile_element) const noexcept
    Definition: distribution.h:247
    dlaf::matrix::Distribution::nextLocalTileFromGlobalTile
    SizeType nextLocalTileFromGlobalTile(SizeType global_tile) const noexcept
    Definition: distribution.h:351
    dlaf::matrix::LayoutInfo
    Definition: layout_info.h:26
    -
    dlaf::matrix::Matrix< const T, D >
    Definition: matrix.h:151
    -
    dlaf::matrix::Matrix
    Definition: matrix.h:55
    +
    dlaf::matrix::Matrix< const T, D >
    Definition: matrix.h:171
    +
    dlaf::matrix::Matrix
    Definition: matrix.h:56
    copy_tile.h
    diff --git a/master/permutations_2general_8h_source.html b/master/permutations_2general_8h_source.html index fd82240d72..affdb84dff 100644 --- a/master/permutations_2general_8h_source.html +++ b/master/permutations_2general_8h_source.html @@ -109,51 +109,59 @@
    52  DLAF_ASSERT(matrix::equal_size(mat_in, mat_out), mat_in);
    53  DLAF_ASSERT(matrix::equal_blocksize(mat_in, mat_out), mat_in);
    54 
    -
    55  DLAF_ASSERT(perms.size().rows() == mat_in.size().rows(), perms, mat_in);
    -
    56  DLAF_ASSERT(perms.size().cols() == 1, perms);
    -
    57  DLAF_ASSERT(perms.blockSize().rows() == mat_in.blockSize().rows(), mat_in, perms);
    +
    55  DLAF_ASSERT(matrix::single_tile_per_block(perms), perms);
    +
    56  DLAF_ASSERT(matrix::single_tile_per_block(mat_in), mat_in);
    +
    57  DLAF_ASSERT(matrix::single_tile_per_block(mat_out), mat_out);
    58 
    -
    59  DLAF_ASSERT(i_begin >= 0 && i_begin <= i_end, i_begin, i_end);
    -
    60  DLAF_ASSERT(i_end <= perms.nrTiles().rows(), i_end, perms);
    -
    61 
    -
    62  internal::Permutations<B, D, T, coord>::call(i_begin, i_end, perms, mat_in, mat_out);
    -
    63 }
    -
    64 
    -
    85 template <Backend B, Device D, class T, Coord coord>
    -
    86 void permute(comm::CommunicatorGrid grid, common::Pipeline<comm::Communicator>& sub_task_chain,
    -
    87  SizeType i_begin, SizeType i_end, Matrix<const SizeType, D>& perms,
    -
    88  Matrix<const T, D>& mat_in, Matrix<T, D>& mat_out) {
    -
    89  DLAF_ASSERT(matrix::local_matrix(perms), perms);
    -
    90  DLAF_ASSERT(matrix::equal_process_grid(mat_in, grid), mat_in, grid);
    -
    91  DLAF_ASSERT(matrix::equal_process_grid(mat_out, grid), mat_out, grid);
    -
    92 
    -
    93  // Note:
    -
    94  // These are not implementation constraints, but more logic constraints. Indeed, these ensure that
    -
    95  // the range [i_begin, i_end] is square in terms of elements (it would not make sense to have it square
    -
    96  // in terms of number of tiles). Moreover, by requiring mat_in and mat_out matrices to have the same
    -
    97  // shape, it is ensured that range [i_begin, i_end] is actually the same on both sides.
    -
    98  DLAF_ASSERT(square_size(mat_in), mat_in);
    -
    99  DLAF_ASSERT(matrix::square_blocksize(mat_in), mat_in);
    -
    100  DLAF_ASSERT(matrix::equal_size(mat_in, mat_out), mat_in);
    -
    101  DLAF_ASSERT(matrix::equal_blocksize(mat_in, mat_out), mat_in);
    -
    102 
    -
    103  DLAF_ASSERT(perms.size().rows() == mat_in.size().rows(), perms, mat_in);
    -
    104  DLAF_ASSERT(perms.size().cols() == 1, perms);
    -
    105  DLAF_ASSERT(perms.blockSize().rows() == mat_in.blockSize().rows(), mat_in, perms);
    +
    59  DLAF_ASSERT(perms.size().rows() == mat_in.size().rows(), perms, mat_in);
    +
    60  DLAF_ASSERT(perms.size().cols() == 1, perms);
    +
    61  DLAF_ASSERT(perms.blockSize().rows() == mat_in.blockSize().rows(), mat_in, perms);
    +
    62 
    +
    63  DLAF_ASSERT(i_begin >= 0 && i_begin <= i_end, i_begin, i_end);
    +
    64  DLAF_ASSERT(i_end <= perms.nrTiles().rows(), i_end, perms);
    +
    65 
    +
    66  internal::Permutations<B, D, T, coord>::call(i_begin, i_end, perms, mat_in, mat_out);
    +
    67 }
    +
    68 
    +
    89 template <Backend B, Device D, class T, Coord coord>
    +
    90 void permute(comm::CommunicatorGrid grid, common::Pipeline<comm::Communicator>& sub_task_chain,
    +
    91  SizeType i_begin, SizeType i_end, Matrix<const SizeType, D>& perms,
    +
    92  Matrix<const T, D>& mat_in, Matrix<T, D>& mat_out) {
    +
    93  DLAF_ASSERT(matrix::local_matrix(perms), perms);
    +
    94  DLAF_ASSERT(matrix::equal_process_grid(mat_in, grid), mat_in, grid);
    +
    95  DLAF_ASSERT(matrix::equal_process_grid(mat_out, grid), mat_out, grid);
    +
    96 
    +
    97  // Note:
    +
    98  // These are not implementation constraints, but more logic constraints. Indeed, these ensure that
    +
    99  // the range [i_begin, i_end] is square in terms of elements (it would not make sense to have it square
    +
    100  // in terms of number of tiles). Moreover, by requiring mat_in and mat_out matrices to have the same
    +
    101  // shape, it is ensured that range [i_begin, i_end] is actually the same on both sides.
    +
    102  DLAF_ASSERT(square_size(mat_in), mat_in);
    +
    103  DLAF_ASSERT(matrix::square_blocksize(mat_in), mat_in);
    +
    104  DLAF_ASSERT(matrix::equal_size(mat_in, mat_out), mat_in);
    +
    105  DLAF_ASSERT(matrix::equal_blocksize(mat_in, mat_out), mat_in);
    106 
    -
    107  DLAF_ASSERT(i_begin >= 0 && i_begin <= i_end, i_begin, i_end);
    -
    108  DLAF_ASSERT(i_end <= perms.nrTiles().rows(), i_end, perms);
    -
    109 
    -
    110  internal::Permutations<B, D, T, coord>::call(sub_task_chain, i_begin, i_end, perms, mat_in, mat_out);
    -
    111 }
    -
    112 
    -
    118 template <Backend B, Device D, class T, Coord coord>
    -
    119 void permute(comm::CommunicatorGrid grid, SizeType i_begin, SizeType i_end,
    -
    120  Matrix<const SizeType, D>& perms, Matrix<const T, D>& mat_in, Matrix<T, D>& mat_out) {
    -
    121  common::Pipeline<comm::Communicator> sub_task_chain(grid.subCommunicator(orthogonal(coord)).clone());
    -
    122  permute<B, D, T, coord>(grid, sub_task_chain, i_begin, i_end, perms, mat_in, mat_out);
    -
    123 }
    -
    124 }
    +
    107  DLAF_ASSERT(matrix::single_tile_per_block(perms), perms);
    +
    108  DLAF_ASSERT(matrix::single_tile_per_block(mat_in), mat_in);
    +
    109  DLAF_ASSERT(matrix::single_tile_per_block(mat_out), mat_out);
    +
    110 
    +
    111  DLAF_ASSERT(perms.size().rows() == mat_in.size().rows(), perms, mat_in);
    +
    112  DLAF_ASSERT(perms.size().cols() == 1, perms);
    +
    113  DLAF_ASSERT(perms.blockSize().rows() == mat_in.blockSize().rows(), mat_in, perms);
    +
    114 
    +
    115  DLAF_ASSERT(i_begin >= 0 && i_begin <= i_end, i_begin, i_end);
    +
    116  DLAF_ASSERT(i_end <= perms.nrTiles().rows(), i_end, perms);
    +
    117 
    +
    118  internal::Permutations<B, D, T, coord>::call(sub_task_chain, i_begin, i_end, perms, mat_in, mat_out);
    +
    119 }
    +
    120 
    +
    126 template <Backend B, Device D, class T, Coord coord>
    +
    127 void permute(comm::CommunicatorGrid grid, SizeType i_begin, SizeType i_end,
    +
    128  Matrix<const SizeType, D>& perms, Matrix<const T, D>& mat_in, Matrix<T, D>& mat_out) {
    +
    129  common::Pipeline<comm::Communicator> sub_task_chain(grid.subCommunicator(orthogonal(coord)).clone());
    +
    130  permute<B, D, T, coord>(grid, sub_task_chain, i_begin, i_end, perms, mat_in, mat_out);
    +
    131 }
    +
    132 }
    dlaf::comm::CommunicatorGrid
    Definition: communicator_grid.h:42
    dlaf::common::Pipeline
    Definition: pipeline.h:31
    communicator_grid.h
    diff --git a/master/print__csv_8h_source.html b/master/print__csv_8h_source.html index 0067cc52c5..2569826fc1 100644 --- a/master/print__csv_8h_source.html +++ b/master/print__csv_8h_source.html @@ -140,7 +140,7 @@
    74 }
    75 }
    -
    dlaf::matrix::Matrix
    Definition: matrix.h:55
    +
    dlaf::matrix::Matrix
    Definition: matrix.h:56
    dlaf::matrix::Tile
    Definition: tile.h:329
    @@ -150,7 +150,7 @@
    dlaf::format::csv
    Definition: print_csv.h:24
    util_matrix.h
    -
    dlaf::matrix::local_matrix
    bool local_matrix(const Matrix< const T, D > &m) noexcept
    Returns true if the matrix is local to a process.
    Definition: util_matrix.h:66
    +
    dlaf::matrix::local_matrix
    bool local_matrix(const Matrix< const T, D > &m) noexcept
    Returns true if the matrix is local to a process.
    Definition: util_matrix.h:72
    diff --git a/master/reduction__to__band_8h.html b/master/reduction__to__band_8h.html index 7ddea9d8c7..11ac78bed0 100644 --- a/master/reduction__to__band_8h.html +++ b/master/reduction__to__band_8h.html @@ -153,6 +153,8 @@

    21 namespace dlaf::eigensolver {
    22 
    26 //
    -
    37 template <Backend B, Device D, class T>
    -
    38 Matrix<T, Device::CPU> reductionToBand(Matrix<T, D>& mat_a, const SizeType band_size) {
    -
    39  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    -
    40  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    -
    41 
    -
    42  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
    +
    38 template <Backend B, Device D, class T>
    +
    39 Matrix<T, Device::CPU> reductionToBand(Matrix<T, D>& mat_a, const SizeType band_size) {
    +
    40  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    +
    41  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    +
    42  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
    43 
    -
    44  DLAF_ASSERT(band_size >= 2, band_size);
    -
    45  DLAF_ASSERT(mat_a.blockSize().rows() % band_size == 0, mat_a.blockSize().rows(), band_size);
    -
    46 
    -
    47  return internal::ReductionToBand<B, D, T>::call(mat_a, band_size);
    -
    48 }
    -
    49 
    -
    66 
    -
    82 
    -
    102 template <Backend B, Device D, class T>
    -
    103 Matrix<T, Device::CPU> reductionToBand(comm::CommunicatorGrid grid, Matrix<T, D>& mat_a,
    -
    104  const SizeType band_size) {
    -
    105  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    -
    106  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    -
    107  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
    -
    108 
    -
    109  DLAF_ASSERT(band_size >= 2, band_size);
    -
    110  DLAF_ASSERT(mat_a.blockSize().rows() % band_size == 0, mat_a.blockSize().rows(), band_size);
    -
    111 
    -
    112  return internal::ReductionToBand<B, D, T>::call(grid, mat_a, band_size);
    -
    113 }
    -
    114 }
    +
    44  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
    +
    45 
    +
    46  DLAF_ASSERT(band_size >= 2, band_size);
    +
    47  DLAF_ASSERT(mat_a.blockSize().rows() % band_size == 0, mat_a.blockSize().rows(), band_size);
    +
    48 
    +
    49  return internal::ReductionToBand<B, D, T>::call(mat_a, band_size);
    +
    50 }
    +
    51 
    +
    68 
    +
    84 
    +
    105 template <Backend B, Device D, class T>
    +
    106 Matrix<T, Device::CPU> reductionToBand(comm::CommunicatorGrid grid, Matrix<T, D>& mat_a,
    +
    107  const SizeType band_size) {
    +
    108  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    +
    109  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    +
    110  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
    +
    111  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
    +
    112 
    +
    113  DLAF_ASSERT(band_size >= 2, band_size);
    +
    114  DLAF_ASSERT(mat_a.blockSize().rows() % band_size == 0, mat_a.blockSize().rows(), band_size);
    +
    115 
    +
    116  return internal::ReductionToBand<B, D, T>::call(grid, mat_a, band_size);
    +
    117 }
    +
    118 }
    dlaf::comm::CommunicatorGrid
    Definition: communicator_grid.h:42
    communicator_grid.h
    -
    dlaf::eigensolver::reductionToBand
    Matrix< T, Device::CPU > reductionToBand(Matrix< T, D > &mat_a, const SizeType band_size)
    Definition: reduction_to_band.h:38
    +
    dlaf::eigensolver::reductionToBand
    Matrix< T, Device::CPU > reductionToBand(Matrix< T, D > &mat_a, const SizeType band_size)
    Definition: reduction_to_band.h:39
    util_matrix.h
    diff --git a/master/retiled__matrix_8h.html b/master/retiled__matrix_8h.html deleted file mode 100644 index 5388d15e3c..0000000000 --- a/master/retiled__matrix_8h.html +++ /dev/null @@ -1,105 +0,0 @@ - - - - - - - -DLAF: /home/runner/work/DLA-Future/DLA-Future/include/dlaf/matrix/retiled_matrix.h File Reference - - - - - - - - - -
    -
    -

     
    file  print_numpy.h [code]
     
    file  retiled_matrix.h [code]
     
    file  tile.h [code]
     
    file  util_distribution.h [code]
     ► memory
     Cdlaf::matrix::Matrix< T, Target >
     ►Cdlaf::matrix::internal::MatrixBase
     Cdlaf::matrix::MatrixMirror< T, Target, Source >
     ►Cdlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >
    - - - - - -
    -
    DLAF -
    -
    - - - - - - - - - -
    -
    - - -
    - -
    - - - -
    -
    -Classes | -Macros
    -
    -
    retiled_matrix.h File Reference
    -
    -
    -
    #include <cstddef>
    -#include <vector>
    -#include <dlaf/common/range2d.h>
    -#include <dlaf/common/vector.h>
    -#include <dlaf/matrix/distribution.h>
    -#include <dlaf/matrix/internal/tile_pipeline.h>
    -#include <dlaf/matrix/matrix.h>
    -#include <dlaf/matrix/matrix_base.h>
    -#include <dlaf/matrix/tile.h>
    -#include <dlaf/types.h>
    -
    -

    Go to the source code of this file.

    - - - - -

    -Classes

    class  dlaf::matrix::RetiledMatrix< T, D >
     
    - - - -

    -Macros

    -#define DLAF_RETILED_MATRIX_ETI(KWORD, DATATYPE, DEVICE)    KWORD template class RetiledMatrix<DATATYPE, DEVICE>;
     
    -
    - -
    -Generated by doxygen 1.9.1 -
    - - diff --git a/master/retiled__matrix_8h_source.html b/master/retiled__matrix_8h_source.html deleted file mode 100644 index 687d255d46..0000000000 --- a/master/retiled__matrix_8h_source.html +++ /dev/null @@ -1,260 +0,0 @@ - - - - - - - -DLAF: /home/runner/work/DLA-Future/DLA-Future/include/dlaf/matrix/retiled_matrix.h Source File - - - - - - - - - -
    -
    - - - - - - -
    -
    DLAF -
    -
    -
    - - - - - - - - -
    -
    - - -
    - -
    - - -
    -
    -
    -
    retiled_matrix.h
    -
    -
    -Go to the documentation of this file.
    1 //
    -
    2 // Distributed Linear Algebra with Future (DLAF)
    -
    3 //
    -
    4 // Copyright (c) 2018-2023, ETH Zurich
    -
    5 // All rights reserved.
    -
    6 //
    -
    7 // Please, refer to the LICENSE file in the root directory.
    -
    8 // SPDX-License-Identifier: BSD-3-Clause
    -
    9 //
    -
    10 
    -
    11 #pragma once
    -
    12 
    -
    14 
    -
    15 #include <cstddef>
    -
    16 #include <vector>
    -
    17 
    -
    18 #include <dlaf/common/range2d.h>
    -
    19 #include <dlaf/common/vector.h>
    -
    20 #include <dlaf/matrix/distribution.h>
    -
    21 #include <dlaf/matrix/internal/tile_pipeline.h>
    -
    22 #include <dlaf/matrix/matrix.h>
    -
    23 #include <dlaf/matrix/matrix_base.h>
    -
    24 #include <dlaf/matrix/tile.h>
    -
    25 #include <dlaf/types.h>
    -
    26 
    -
    27 namespace dlaf::matrix {
    -
    28 
    -
    35 template <class T, Device D>
    -
    36 class RetiledMatrix : public internal::MatrixBase {
    -
    37 public:
    -
    38  static constexpr Device device = D;
    -
    39 
    -
    40  using ElementType = T;
    -
    41  using TileType = Tile<ElementType, D>;
    -
    42  using ConstTileType = Tile<const ElementType, D>;
    -
    43  using TileDataType = internal::TileData<const ElementType, D>;
    -
    44 
    -
    46  RetiledMatrix(Matrix<T, D>& mat, const LocalTileSize& tiles_per_block)
    -
    47  : MatrixBase(mat.distribution(), tiles_per_block) {
    -
    48  using common::internal::vector;
    -
    49  namespace ex = pika::execution::experimental;
    -
    50 
    -
    51  const auto n = to_sizet(distribution().localNrTiles().linear_size());
    -
    52  tile_managers_.reserve(n);
    -
    53  for (std::size_t i = 0; i < n; ++i) {
    -
    54  tile_managers_.emplace_back(Tile<T, D>());
    -
    55  }
    -
    56 
    -
    57  const auto tile_size = distribution().baseTileSize();
    -
    58  vector<SubTileSpec> specs;
    -
    59  vector<LocalTileIndex> indices;
    -
    60  specs.reserve(tiles_per_block.linear_size());
    -
    61  indices.reserve(tiles_per_block.linear_size());
    -
    62 
    -
    63  for (const auto& orig_tile_index : common::iterate_range2d(mat.distribution().localNrTiles())) {
    -
    64  const auto original_tile_size = mat.tileSize(mat.distribution().globalTileIndex(orig_tile_index));
    -
    65 
    -
    66  for (SizeType j = 0; j < original_tile_size.cols(); j += tile_size.cols())
    -
    67  for (SizeType i = 0; i < original_tile_size.rows(); i += tile_size.rows()) {
    -
    68  indices.emplace_back(
    -
    69  LocalTileIndex{orig_tile_index.row() * tiles_per_block.rows() + i / tile_size.rows(),
    -
    70  orig_tile_index.col() * tiles_per_block.cols() + j / tile_size.cols()});
    -
    71  specs.emplace_back(SubTileSpec{{i, j},
    -
    72  tileSize(distribution().globalTileIndex(indices.back()))});
    -
    73  }
    -
    74 
    -
    75  auto sub_tiles = splitTileDisjoint(mat.readwrite(orig_tile_index), specs);
    -
    76 
    -
    77  DLAF_ASSERT_HEAVY(specs.size() == indices.size(), specs.size(), indices.size());
    -
    78  for (SizeType j = 0; j < specs.size(); ++j) {
    -
    79  const auto i = tileLinearIndex(indices[j]);
    -
    80 
    -
    81  // Move subtile to be managed by the tile manager of RetiledMatrix. We
    -
    82  // use readwrite_with_wrapper to get access to the original tile managed
    -
    83  // by the underlying async_rw_mutex.
    -
    84  auto s =
    -
    85  ex::when_all(tile_managers_[i].readwrite_with_wrapper(), std::move(sub_tiles[to_sizet(j)])) |
    -
    86  ex::then([](internal::TileAsyncRwMutexReadWriteWrapper<T, D> empty_tile_wrapper,
    -
    87  Tile<T, D> sub_tile) { empty_tile_wrapper.get() = std::move(sub_tile); });
    -
    88  ex::start_detached(std::move(s));
    -
    89  }
    -
    90 
    -
    91  specs.clear();
    -
    92  indices.clear();
    -
    93  }
    -
    94  }
    -
    95 
    -
    96  RetiledMatrix(const RetiledMatrix& rhs) = delete;
    -
    97  RetiledMatrix(RetiledMatrix&& rhs) = default;
    -
    98 
    -
    99  RetiledMatrix& operator=(const RetiledMatrix& rhs) = delete;
    -
    100  RetiledMatrix& operator=(RetiledMatrix&& rhs) = default;
    -
    101 
    -
    105  ReadOnlyTileSender<T, D> read(const LocalTileIndex& index) noexcept {
    -
    106  const auto i = tileLinearIndex(index);
    -
    107  return tile_managers_[i].read();
    -
    108  }
    -
    109 
    -
    114  ReadOnlyTileSender<T, D> read(const GlobalTileIndex& index) noexcept {
    -
    115  return read(distribution().localTileIndex(index));
    -
    116  }
    -
    117 
    -
    122  void waitLocalTiles() noexcept {
    -
    123  auto readwrite_f = [this](const LocalTileIndex& index) {
    -
    124  return this->tile_managers_[tileLinearIndex(index)].readwrite();
    -
    125  };
    -
    126 
    -
    127  const auto range_local = common::iterate_range2d(distribution().localNrTiles());
    -
    128  pika::this_thread::experimental::sync_wait(pika::execution::experimental::when_all_vector(
    -
    129  internal::selectGeneric(readwrite_f, range_local)));
    -
    130  }
    -
    131 
    -
    135  ReadWriteTileSender<T, D> readwrite(const LocalTileIndex& index) noexcept {
    -
    136  const auto i = tileLinearIndex(index);
    -
    137  return tile_managers_[i].readwrite();
    -
    138  }
    -
    139 
    -
    144  ReadWriteTileSender<T, D> readwrite(const GlobalTileIndex& index) noexcept {
    -
    145  return readwrite(distribution().localTileIndex(index));
    -
    146  }
    -
    147 
    -
    148  void done(const LocalTileIndex& index) noexcept {
    -
    149  const auto i = tileLinearIndex(index);
    -
    150  tile_managers_[i].reset();
    -
    151  }
    -
    152 
    -
    153  void done(const GlobalTileIndex& index) noexcept {
    -
    154  done(distribution().localTileIndex(index));
    -
    155  }
    -
    156 
    -
    157 protected:
    -
    158  std::vector<internal::TilePipeline<T, D>> tile_managers_;
    -
    159 };
    -
    160 
    -
    162 template <class T, Device D>
    -
    163 class RetiledMatrix<const T, D>;
    -
    164 
    -
    165 // ETI
    -
    166 
    -
    167 #define DLAF_RETILED_MATRIX_ETI(KWORD, DATATYPE, DEVICE) \
    -
    168  KWORD template class RetiledMatrix<DATATYPE, DEVICE>;
    -
    169 
    -
    170 DLAF_RETILED_MATRIX_ETI(extern, float, Device::CPU)
    -
    171 DLAF_RETILED_MATRIX_ETI(extern, double, Device::CPU)
    -
    172 DLAF_RETILED_MATRIX_ETI(extern, std::complex<float>, Device::CPU)
    -
    173 DLAF_RETILED_MATRIX_ETI(extern, std::complex<double>, Device::CPU)
    -
    174 
    -
    175 #if defined(DLAF_WITH_GPU)
    -
    176 DLAF_RETILED_MATRIX_ETI(extern, float, Device::GPU)
    -
    177 DLAF_RETILED_MATRIX_ETI(extern, double, Device::GPU)
    -
    178 DLAF_RETILED_MATRIX_ETI(extern, std::complex<float>, Device::GPU)
    -
    179 DLAF_RETILED_MATRIX_ETI(extern, std::complex<double>, Device::GPU)
    -
    180 #endif
    -
    181 }
    - - -
    dlaf::matrix::Matrix
    Definition: matrix.h:55
    -
    dlaf::matrix::Matrix::readwrite
    ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
    Definition: matrix.h:121
    -
    dlaf::matrix::RetiledMatrix
    Definition: retiled_matrix.h:36
    -
    dlaf::matrix::RetiledMatrix::RetiledMatrix
    RetiledMatrix(Matrix< T, D > &mat, const LocalTileSize &tiles_per_block)
    Definition: retiled_matrix.h:46
    -
    dlaf::matrix::RetiledMatrix::read
    ReadOnlyTileSender< T, D > read(const LocalTileIndex &index) noexcept
    Definition: retiled_matrix.h:105
    -
    dlaf::matrix::RetiledMatrix::readwrite
    ReadWriteTileSender< T, D > readwrite(const LocalTileIndex &index) noexcept
    Definition: retiled_matrix.h:135
    -
    dlaf::matrix::RetiledMatrix::waitLocalTiles
    void waitLocalTiles() noexcept
    Definition: retiled_matrix.h:122
    -
    dlaf::matrix::RetiledMatrix::readwrite
    ReadWriteTileSender< T, D > readwrite(const GlobalTileIndex &index) noexcept
    Definition: retiled_matrix.h:144
    -
    dlaf::matrix::RetiledMatrix::read
    ReadOnlyTileSender< T, D > read(const GlobalTileIndex &index) noexcept
    Definition: retiled_matrix.h:114
    -
    dlaf::matrix::Tile
    Definition: tile.h:329
    -
    dlaf::matrix::internal::MatrixBase
    Definition: matrix_base.h:24
    -
    dlaf::matrix::internal::MatrixBase::distribution
    const matrix::Distribution & distribution() const noexcept
    Returns the distribution of the matrix.
    Definition: matrix_base.h:90
    -
    dlaf::matrix::internal::MatrixBase::tileLinearIndex
    std::size_t tileLinearIndex(const LocalTileIndex &index) const noexcept
    Definition: matrix_base.h:111
    -
    dlaf::matrix::internal::MatrixBase::tileSize
    TileElementSize tileSize(const GlobalTileIndex &index) const noexcept
    Returns the size of the Tile with global index index.
    Definition: matrix_base.h:95
    - - - -
    dlaf::matrix::splitTileDisjoint
    std::vector< ReadWriteTileSender< T, D > > splitTileDisjoint(ReadWriteTileSender< T, D > &&tile, const std::vector< SubTileSpec > &specs)
    Definition: tile.h:570
    - -
    matrix_base.h
    - -
    dlaf::common::internal::vector
    Definition: vector.h:27
    -
    dlaf::matrix::SubTileSpec
    Contains the information to create a subtile.
    Definition: tile.h:109
    - -
    dlaf::to_sizet
    auto to_sizet(const T signed_value)
    Definition: types.h:265
    - -
    - -
    -Generated by doxygen 1.9.1 -
    - - diff --git a/master/rot_8h_source.html b/master/rot_8h_source.html index 930ce43641..f778e02f20 100644 --- a/master/rot_8h_source.html +++ b/master/rot_8h_source.html @@ -448,7 +448,7 @@ -
    dlaf::matrix::select
    auto select(MatrixLike &matrix, common::IterableRange2D< SizeType, LocalTile_TAG > range)
    Definition: matrix.h:392
    +
    dlaf::matrix::select
    auto select(MatrixLike &matrix, common::IterableRange2D< SizeType, LocalTile_TAG > range)
    Definition: matrix.h:449
    memory_view.h
    diff --git a/master/search/all_10.js b/master/search/all_10.js index 91fbc31ca4..7fad8e0000 100644 --- a/master/search/all_10.js +++ b/master/search/all_10.js @@ -1,34 +1,34 @@ var searchData= [ - ['range2d_2eh_343',['range2d.h',['../range2d_8h.html',1,'']]], - ['rangeend_344',['rangeEnd',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a3aa72e4de5ae6c2d87b0643e08311df8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['rangeendlocal_345',['rangeEndLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a7aa91d49db2332c93e5b2b2e1aadbecf',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['rangestart_346',['rangeStart',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a09572f6a6461327f0ed6d0e471174410',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['rangestartlocal_347',['rangeStartLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#acef133d14867a04758e8914135c5d5f8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['rank_348',['rank',['../classdlaf_1_1comm_1_1_communicator_grid.html#a45574d05dd3cfb3d40d198cdf10b169e',1,'dlaf::comm::CommunicatorGrid::rank()'],['../classdlaf_1_1comm_1_1_communicator.html#a52be17b02755dcbfbdcaa95ba5f9230c',1,'dlaf::comm::Communicator::rank()']]], - ['rankfullcommunicator_349',['rankFullCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#ad9a0a0a05723adbe0e57171c7c1e2e93',1,'dlaf::comm::CommunicatorGrid']]], - ['rankglobalelement_350',['rankGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a8eac06b3be1805d7261a75809529a910',1,'dlaf::matrix::Distribution']]], - ['rankglobaltile_351',['rankGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#a49ff769692620e7f94cf796a65ff6a9e',1,'dlaf::matrix::Distribution::rankGlobalTile()'],['../util__distribution_8h.html#ad89b3802ea0ed39625419b53e3f199d1',1,'dlaf::util::matrix::rankGlobalTile()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a622f4764a792986112c89f282e04080f',1,'dlaf::matrix::internal::MatrixBase::rankGlobalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#ae3f7cda513d213773cb4d8a0ba22b758',1,'dlaf::matrix::Distribution::rankGlobalTile()']]], - ['rankindex_352',['rankIndex',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a685f321909fc853e023bff712c964c2d',1,'dlaf::matrix::internal::MatrixBase']]], - ['rdma_2eh_353',['rdma.h',['../rdma_8h.html',1,'']]], - ['read_354',['read',['../classdlaf_1_1matrix_1_1_retiled_matrix.html#afacf9af5a45a341c284844d866227579',1,'dlaf::matrix::RetiledMatrix::read(const GlobalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#a814972188382d0ae0120b6acfddd92bb',1,'dlaf::matrix::RetiledMatrix::read(const LocalTileIndex &index) noexcept'],['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a4631ad4a6fa72e74cf95798c527a48d1',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::read()'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#a8485d802a7f318a845de46e1b4e48118',1,'dlaf::matrix::Matrix< const T, D >::read(const GlobalTileIndex &index)'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#abd8114d7c9ff03a2ff5c8613284135f6',1,'dlaf::matrix::Matrix< const T, D >::read(const LocalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a99845f6d6b238d3a8d543d20d3adb04d',1,'dlaf::matrix::internal::TilePipeline::read()']]], - ['readwrite_355',['readwrite',['../classdlaf_1_1matrix_1_1_matrix.html#ad11e9036fe317b6328ae71585b19c811',1,'dlaf::matrix::Matrix::readwrite()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#ab0a26ee4322ab4b7a3d250a82d072f68',1,'dlaf::matrix::internal::TilePipeline::readwrite()'],['../classdlaf_1_1matrix_1_1_matrix.html#af486e66243b57df94b9822bd2553ae38',1,'dlaf::matrix::Matrix::readwrite()'],['../structdlaf_1_1matrix_1_1_panel.html#aac41b4dfa833e4a24839fe1d470b835a',1,'dlaf::matrix::Panel::readwrite()'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#aa0e45d9a97f8d3ff418d2bd08211a5bf',1,'dlaf::matrix::RetiledMatrix::readwrite(const LocalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#ac3ca2e52e6627f5890bcb25fcff60167',1,'dlaf::matrix::RetiledMatrix::readwrite(const GlobalTileIndex &index) noexcept']]], - ['readwrite_5fwith_5fwrapper_356',['readwrite_with_wrapper',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a17403152214a88994a75395f7bebfe3c',1,'dlaf::matrix::internal::TilePipeline']]], - ['receive_5ffrom_357',['receive_from',['../sync_2broadcast_8h.html#a86fb5d23e8299cbd380a4e7d724c1112',1,'dlaf::comm::sync::broadcast::receive_from()'],['../basic_8h.html#a13040096b22855cefca83ec02f0c64c0',1,'dlaf::comm::sync::receive_from(int sender_rank, Communicator &communicator, DataOut &&data)']]], - ['reduce_358',['reduce',['../sync_2reduce_8h.html#a22dfe268f1dcd2f08e2678af288063be',1,'dlaf::comm::sync']]], - ['reduce_2eh_359',['reduce.h',['../kernels_2reduce_8h.html',1,'(Global Namespace)'],['../sync_2reduce_8h.html',1,'(Global Namespace)']]], - ['reduceinplace_360',['reduceInPlace',['../sync_2reduce_8h.html#a86b20000f6542c605c6b5f2bcc9e625b',1,'dlaf::comm::sync']]], - ['reducerecv_361',['reduceRecv',['../sync_2reduce_8h.html#a03e9b0551a974dfa12fc9885272a1bee',1,'dlaf::comm::sync']]], - ['reducerecvinplace_362',['reduceRecvInPlace',['../sync_2reduce_8h.html#a1ce87e8bcfccf05e7edd9d23e1f778f3',1,'dlaf::comm::sync']]], - ['reducesend_363',['reduceSend',['../sync_2reduce_8h.html#a27df5989f95087d0259a3e381b1539b3',1,'dlaf::comm::sync']]], - ['reduction_5fto_5fband_2eh_364',['reduction_to_band.h',['../reduction__to__band_8h.html',1,'']]], - ['reductiontoband_365',['ReductionToBand',['../structdlaf_1_1eigensolver_1_1internal_1_1_reduction_to_band.html',1,'dlaf::eigensolver::internal']]], - ['reductiontoband_366',['reductionToBand',['../reduction__to__band_8h.html#a00574f61da45042c5276b043d7c33617',1,'dlaf::eigensolver::reductionToBand(Matrix< T, D > &mat_a, const SizeType band_size)'],['../reduction__to__band_8h.html#a143b74447a1f72d4d5368b9bad10b331',1,'dlaf::eigensolver::reductionToBand(comm::CommunicatorGrid grid, Matrix< T, D > &mat_a, const SizeType band_size)']]], - ['reset_367',['reset',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#acb87326419c25da556f027565b96c086',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::reset()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a83ffbae7e4944baa1bc4bfbe5dd1100d',1,'dlaf::matrix::internal::TilePipeline::reset()'],['../classdlaf_1_1common_1_1_pipeline.html#a4a9e1a35ffad351b5b4e01b88ef8a447',1,'dlaf::common::Pipeline::reset()']]], - ['retiled_5fmatrix_2eh_368',['retiled_matrix.h',['../retiled__matrix_8h.html',1,'']]], - ['retiledmatrix_369',['RetiledMatrix',['../classdlaf_1_1matrix_1_1_retiled_matrix.html',1,'dlaf::matrix::RetiledMatrix< T, D >'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#a4e208b9159a478b3fa20167ec8e04318',1,'dlaf::matrix::RetiledMatrix::RetiledMatrix()']]], - ['round_5frobin_2eh_370',['round_robin.h',['../round__robin_8h.html',1,'']]], - ['roundrobin_371',['RoundRobin',['../structdlaf_1_1common_1_1_round_robin.html',1,'dlaf::common']]], - ['roundrobin_3c_20dlaf_3a_3amatrix_3a_3apanel_3c_20coord_3a_3acol_2c_20t_2c_20device_3a_3acpu_20_3e_20_3e_372',['RoundRobin< dlaf::matrix::Panel< Coord::Col, T, Device::CPU > >',['../structdlaf_1_1common_1_1_round_robin.html',1,'dlaf::common']]], - ['rowcommunicator_373',['rowCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#ab13487adb2992d2cc50e06b498582391',1,'dlaf::comm::CommunicatorGrid']]] + ['range2d_2eh_344',['range2d.h',['../range2d_8h.html',1,'']]], + ['rangeend_345',['rangeEnd',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a3aa72e4de5ae6c2d87b0643e08311df8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['rangeendlocal_346',['rangeEndLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a7aa91d49db2332c93e5b2b2e1aadbecf',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['rangestart_347',['rangeStart',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a09572f6a6461327f0ed6d0e471174410',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['rangestartlocal_348',['rangeStartLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#acef133d14867a04758e8914135c5d5f8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['rank_349',['rank',['../classdlaf_1_1comm_1_1_communicator_grid.html#a45574d05dd3cfb3d40d198cdf10b169e',1,'dlaf::comm::CommunicatorGrid::rank()'],['../classdlaf_1_1comm_1_1_communicator.html#a52be17b02755dcbfbdcaa95ba5f9230c',1,'dlaf::comm::Communicator::rank()']]], + ['rankfullcommunicator_350',['rankFullCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#ad9a0a0a05723adbe0e57171c7c1e2e93',1,'dlaf::comm::CommunicatorGrid']]], + ['rankglobalelement_351',['rankGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a8eac06b3be1805d7261a75809529a910',1,'dlaf::matrix::Distribution']]], + ['rankglobaltile_352',['rankGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#a49ff769692620e7f94cf796a65ff6a9e',1,'dlaf::matrix::Distribution::rankGlobalTile()'],['../util__distribution_8h.html#ad89b3802ea0ed39625419b53e3f199d1',1,'dlaf::util::matrix::rankGlobalTile()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a622f4764a792986112c89f282e04080f',1,'dlaf::matrix::internal::MatrixBase::rankGlobalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#ae3f7cda513d213773cb4d8a0ba22b758',1,'dlaf::matrix::Distribution::rankGlobalTile()']]], + ['rankindex_353',['rankIndex',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a685f321909fc853e023bff712c964c2d',1,'dlaf::matrix::internal::MatrixBase']]], + ['rdma_2eh_354',['rdma.h',['../rdma_8h.html',1,'']]], + ['read_355',['read',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a4631ad4a6fa72e74cf95798c527a48d1',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::read()'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#a8485d802a7f318a845de46e1b4e48118',1,'dlaf::matrix::Matrix< const T, D >::read()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a99845f6d6b238d3a8d543d20d3adb04d',1,'dlaf::matrix::internal::TilePipeline::read()'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#abd8114d7c9ff03a2ff5c8613284135f6',1,'dlaf::matrix::Matrix< const T, D >::read()']]], + ['readwrite_356',['readwrite',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#ab0a26ee4322ab4b7a3d250a82d072f68',1,'dlaf::matrix::internal::TilePipeline::readwrite()'],['../structdlaf_1_1matrix_1_1_panel.html#aac41b4dfa833e4a24839fe1d470b835a',1,'dlaf::matrix::Panel::readwrite()'],['../classdlaf_1_1matrix_1_1_matrix.html#af486e66243b57df94b9822bd2553ae38',1,'dlaf::matrix::Matrix::readwrite(const GlobalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#ad11e9036fe317b6328ae71585b19c811',1,'dlaf::matrix::Matrix::readwrite(const LocalTileIndex &index) noexcept']]], + ['readwrite_5fwith_5fwrapper_357',['readwrite_with_wrapper',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a17403152214a88994a75395f7bebfe3c',1,'dlaf::matrix::internal::TilePipeline']]], + ['receive_5ffrom_358',['receive_from',['../sync_2broadcast_8h.html#a86fb5d23e8299cbd380a4e7d724c1112',1,'dlaf::comm::sync::broadcast::receive_from()'],['../basic_8h.html#a13040096b22855cefca83ec02f0c64c0',1,'dlaf::comm::sync::receive_from(int sender_rank, Communicator &communicator, DataOut &&data)']]], + ['reduce_359',['reduce',['../sync_2reduce_8h.html#a22dfe268f1dcd2f08e2678af288063be',1,'dlaf::comm::sync']]], + ['reduce_2eh_360',['reduce.h',['../kernels_2reduce_8h.html',1,'(Global Namespace)'],['../sync_2reduce_8h.html',1,'(Global Namespace)']]], + ['reduceinplace_361',['reduceInPlace',['../sync_2reduce_8h.html#a86b20000f6542c605c6b5f2bcc9e625b',1,'dlaf::comm::sync']]], + ['reducerecv_362',['reduceRecv',['../sync_2reduce_8h.html#a03e9b0551a974dfa12fc9885272a1bee',1,'dlaf::comm::sync']]], + ['reducerecvinplace_363',['reduceRecvInPlace',['../sync_2reduce_8h.html#a1ce87e8bcfccf05e7edd9d23e1f778f3',1,'dlaf::comm::sync']]], + ['reducesend_364',['reduceSend',['../sync_2reduce_8h.html#a27df5989f95087d0259a3e381b1539b3',1,'dlaf::comm::sync']]], + ['reduction_5fto_5fband_2eh_365',['reduction_to_band.h',['../reduction__to__band_8h.html',1,'']]], + ['reductiontoband_366',['ReductionToBand',['../structdlaf_1_1eigensolver_1_1internal_1_1_reduction_to_band.html',1,'dlaf::eigensolver::internal']]], + ['reductiontoband_367',['reductionToBand',['../reduction__to__band_8h.html#a00574f61da45042c5276b043d7c33617',1,'dlaf::eigensolver::reductionToBand(Matrix< T, D > &mat_a, const SizeType band_size)'],['../reduction__to__band_8h.html#a143b74447a1f72d4d5368b9bad10b331',1,'dlaf::eigensolver::reductionToBand(comm::CommunicatorGrid grid, Matrix< T, D > &mat_a, const SizeType band_size)']]], + ['reset_368',['reset',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#acb87326419c25da556f027565b96c086',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::reset()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a83ffbae7e4944baa1bc4bfbe5dd1100d',1,'dlaf::matrix::internal::TilePipeline::reset()'],['../classdlaf_1_1common_1_1_pipeline.html#a4a9e1a35ffad351b5b4e01b88ef8a447',1,'dlaf::common::Pipeline::reset()']]], + ['retiledsubpipeline_369',['retiledSubPipeline',['../classdlaf_1_1matrix_1_1_matrix.html#aae1c8b22f9862944a006003f528595ff',1,'dlaf::matrix::Matrix']]], + ['retiledsubpipelineconst_370',['retiledSubPipelineConst',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#aa10512ea4bde924a55e73bfe07050942',1,'dlaf::matrix::Matrix< const T, D >']]], + ['round_5frobin_2eh_371',['round_robin.h',['../round__robin_8h.html',1,'']]], + ['roundrobin_372',['RoundRobin',['../structdlaf_1_1common_1_1_round_robin.html',1,'dlaf::common']]], + ['roundrobin_3c_20dlaf_3a_3amatrix_3a_3apanel_3c_20coord_3a_3acol_2c_20t_2c_20device_3a_3acpu_20_3e_20_3e_373',['RoundRobin< dlaf::matrix::Panel< Coord::Col, T, Device::CPU > >',['../structdlaf_1_1common_1_1_round_robin.html',1,'dlaf::common']]], + ['rowcommunicator_374',['rowCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#ab13487adb2992d2cc50e06b498582391',1,'dlaf::comm::CommunicatorGrid']]] ]; diff --git a/master/search/all_11.js b/master/search/all_11.js index 7b677b0825..2da024476b 100644 --- a/master/search/all_11.js +++ b/master/search/all_11.js @@ -1,68 +1,71 @@ var searchData= [ - ['samesign_374',['sameSign',['../util__math_8h.html#a0922553e7cd5d09cc857bb33e8a1b91b',1,'dlaf::util']]], - ['scal_2eh_375',['scal.h',['../scal_8h.html',1,'']]], - ['scheduleallreduce_376',['scheduleAllReduce',['../kernels_2all__reduce_8h.html#a8cb64ca1859aef0b95cdcfa8a1f05297',1,'dlaf::comm']]], - ['scheduleallreduceinplace_377',['scheduleAllReduceInPlace',['../kernels_2all__reduce_8h.html#a689b3d06eea6e5d2c9899521e970ecb0',1,'dlaf::comm']]], - ['scheduleallsump2p_378',['scheduleAllSumP2P',['../p2p__allsum_8h.html#ac786ec776d72b61529c8a79c5baffff4',1,'dlaf::comm']]], - ['schedulerecvbcast_379',['scheduleRecvBcast',['../kernels_2broadcast_8h.html#a3645d0cfc3caad43c37c355896a7215f',1,'dlaf::comm']]], - ['schedulereducerecvinplace_380',['scheduleReduceRecvInPlace',['../kernels_2reduce_8h.html#a89af8ab199d6ba8902cc962122477f00',1,'dlaf::comm']]], - ['schedulereducesend_381',['scheduleReduceSend',['../kernels_2reduce_8h.html#a60899e3a349b0defe79b1561045a373a',1,'dlaf::comm']]], - ['schedulers_2eh_382',['schedulers.h',['../schedulers_8h.html',1,'']]], - ['schedulesendbcast_383',['scheduleSendBcast',['../kernels_2broadcast_8h.html#a98b583f24caaccdf37f3db1f3b46b457',1,'dlaf::comm']]], - ['scopedinitializer_384',['ScopedInitializer',['../structdlaf_1_1_scoped_initializer.html',1,'dlaf']]], - ['select_385',['select',['../matrix_8h.html#a1abc47c06dc1c46c9bdc2425090a02ab',1,'dlaf::matrix']]], - ['selectgeneric_386',['selectGeneric',['../matrix_8h.html#ac3737882aaaf23b1f901da6396202ba1',1,'dlaf::matrix::internal']]], - ['selectread_387',['selectRead',['../matrix_8h.html#a02db5399b6bc7e53a162f5939ed7ca20',1,'dlaf::matrix']]], - ['send_388',['send',['../sync_2broadcast_8h.html#a313a9dc3d212f70d7af273fa0c681aee',1,'dlaf::comm::sync::broadcast']]], - ['send_5fto_389',['send_to',['../basic_8h.html#adfd3205605f800afb2208146565a3bd4',1,'dlaf::comm::sync']]], - ['sendersinglevaluetypeimpl_390',['SenderSingleValueTypeImpl',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl.html',1,'dlaf::internal']]], - ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20rwtype_2c_20rtype_2c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5ftype_3a_3aread_20_3e_20_3e_20_3e_20_3e_391',['SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::read > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01pika_356df456d49829a554002eacb85cc5b7.html',1,'dlaf::internal']]], - ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20rwtype_2c_20rtype_2c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5ftype_3a_3areadwrite_20_3e_20_3e_20_3e_20_3e_392',['SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::readwrite > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01pika_3535bbe9d77284105a6158ec315ee38c.html',1,'dlaf::internal']]], - ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20std_3a_3areference_5fwrapper_3c_20t_20_3e_20_3e_20_3e_20_3e_393',['SenderSingleValueTypeImpl< TypeList< TypeList< std::reference_wrapper< T > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01std_1b569951bd408b620bb545799a6089d25.html',1,'dlaf::internal']]], - ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20t_20_3e_20_3e_20_3e_394',['SenderSingleValueTypeImpl< TypeList< TypeList< T > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01_t_01_4_01_4_01_4.html',1,'dlaf::internal']]], - ['set_395',['set',['../util__matrix_8h.html#a1f796f1fd4dd705c90ac141d687843b2',1,'dlaf::matrix::util::set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f)'],['../util__matrix_8h.html#af07478cd4a633a1422e160863d3d8510',1,'dlaf::matrix::util::set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f, const blas::Op op)']]], - ['set0_396',['set0',['../lapack_2tile_8h.html#ab094eeac6052422b50a298b42a947251',1,'dlaf::tile::set0()'],['../util__matrix_8h.html#a6cdf6afc18a6456ed68b3fd1dffa9c39',1,'dlaf::matrix::util::set0(pika::execution::thread_priority priority, LocalTileIndex begin, LocalTileSize sz, Matrix< T, D > &matrix)'],['../util__matrix_8h.html#a79688b3669b67c143114cf2eef3ba69b',1,'dlaf::matrix::util::set0(pika::execution::thread_priority priority, Panel< axis, T, D, storage > &panel)']]], - ['set_5frandom_397',['set_random',['../util__matrix_8h.html#a8e8c0d0551fbda5968d6c84eb3c70fc0',1,'dlaf::matrix::util']]], - ['set_5frandom_5fhermitian_398',['set_random_hermitian',['../util__matrix_8h.html#a3f7866203ad62c615d11ca18af4d4476',1,'dlaf::matrix::util']]], - ['set_5frandom_5fhermitian_5fpositive_5fdefinite_399',['set_random_hermitian_positive_definite',['../util__matrix_8h.html#ae2afbf7cdec8225e544b7b71aa52ee58',1,'dlaf::matrix::util']]], - ['set_5frandom_5fhermitian_5fwith_5foffset_400',['set_random_hermitian_with_offset',['../util__matrix_8h.html#a2f1823d0004b53105c4d4b7373208133',1,'dlaf::matrix::util::internal']]], - ['setheight_401',['setHeight',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a5ffd11dcc29ca2731c0cf232cb599d8e',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setrange_402',['setRange',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a88f387bed0a5cf8dabd5f6c5b3d2c3b0',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setrangeend_403',['setRangeEnd',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a542319871ee2379e84651d7dd91f0749',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setrangestart_404',['setRangeStart',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a621d6ffd26cb962792c4087159df2769',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['settile_405',['setTile',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa9c0ce49a7829488736d99011c5ac6c9',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setupinternalmatrix_406',['setupInternalMatrix',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a11f5026c35f76adb7c99c725dc969c44',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setwidth_407',['setWidth',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a74e5d5d87edfb1d2de8a54cfa812bd04',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['sharereadwritetile_408',['shareReadWriteTile',['../matrix_2tile_8h.html#aa8e4a64946c725cef06834fb80b75276',1,'dlaf::matrix']]], - ['single_5fthreaded_5fblas_2eh_409',['single_threaded_blas.h',['../single__threaded__blas_8h.html',1,'']]], - ['singlethreadedblasscope_410',['SingleThreadedBlasScope',['../classdlaf_1_1common_1_1internal_1_1_single_threaded_blas_scope.html',1,'dlaf::common::internal']]], - ['size_411',['size',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#adc5a25970dab7cdea07e2f98328df805',1,'dlaf::matrix::internal::MatrixBase::size()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a7ef713ad7d24caccfc0863ab4b8481f2',1,'dlaf::matrix::Tile< const T, D >::size()'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a886c87812f0a8eb9071c70477ac9f042',1,'dlaf::memory::MemoryChunk::size()'],['../classdlaf_1_1memory_1_1_memory_view.html#a5ae9491601035e4f2fd87352916f84c2',1,'dlaf::memory::MemoryView::size()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#a9307e8d79bffad58211c317c92c7dc77',1,'dlaf::comm::CommunicatorGrid::size()'],['../classdlaf_1_1comm_1_1_communicator.html#a415f5d36c3033320074df329163c3127',1,'dlaf::comm::Communicator::size()']]], - ['size2d_412',['Size2D',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common::Size2D< IndexT, Tag >'],['../communicator__grid_8h.html#a1af44f2ea119a10f873e64cf048ba94a',1,'dlaf::comm::Size2D()']]], - ['size2d_3c_20indext_5fmpi_2c_20tag_5fmpi_20_3e_413',['Size2D< IndexT_MPI, TAG_MPI >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3aglobalelement_5ftag_20_3e_414',['Size2D< SizeType, matrix::GlobalElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3aglobaltile_5ftag_20_3e_415',['Size2D< SizeType, matrix::GlobalTile_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3alocalelement_5ftag_20_3e_416',['Size2D< SizeType, matrix::LocalElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3alocaltile_5ftag_20_3e_417',['Size2D< SizeType, matrix::LocalTile_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3atileelement_5ftag_20_3e_418',['Size2D< SizeType, matrix::TileElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['solver_2eh_419',['solver.h',['../solver_8h.html',1,'']]], - ['source_5flocation_420',['source_location',['../structdlaf_1_1common_1_1internal_1_1source__location.html',1,'dlaf::common::internal']]], - ['source_5flocation_421',['SOURCE_LOCATION',['../source__location_8h.html#aa6e6fc9684bd84ce7b72da6a2eef668f',1,'source_location.h']]], - ['source_5flocation_2eh_422',['source_location.h',['../source__location_8h.html',1,'']]], - ['splittile_423',['splitTile',['../matrix_2tile_8h.html#ae5697bd5c424a95890978dfdc4739bf6',1,'dlaf::matrix::splitTile(ReadWriteTileSender< T, D > &&tile, const SubTileSpec &spec)'],['../matrix_2tile_8h.html#a7a2383f364417fce38d6e7038cb18bd1',1,'dlaf::matrix::splitTile(ReadOnlyTileSender< T, D > tile, const std::vector< SubTileSpec > &specs)'],['../matrix_2tile_8h.html#ae46de76ccc35985387f6820b1db9fb51',1,'dlaf::matrix::splitTile(ReadOnlyTileSender< T, D > tile, const SubTileSpec &spec)']]], - ['splittiledisjoint_424',['splitTileDisjoint',['../matrix_2tile_8h.html#a4c15e2e57d6336347e6e19d945bd650e',1,'dlaf::matrix']]], - ['square_5fblocksize_425',['square_blocksize',['../util__matrix_8h.html#a355682eaf279129f2aa06c8e96eaa61b',1,'dlaf::matrix']]], - ['square_5fsize_426',['square_size',['../util__matrix_8h.html#aae6464887dae32e01668a32e71397cb0',1,'dlaf::matrix::square_size()'],['../util__tile_8h.html#adfd1e0b37d590f9f6b56cdb276688f0f',1,'dlaf::tile::square_size()']]], - ['start_5f_427',['start_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ac49583acc02b6ba394d9b4e3b7450e53',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['start_5foffset_5f_428',['start_offset_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a8603ffce77527952a466a9630d98cd5c',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['stedc_429',['stedc',['../lapack_2tile_8h.html#a68163899b7b15f447fdbb05e475d9cc6',1,'dlaf::tile']]], - ['stride_430',['stride',['../structdlaf_1_1common_1_1_data_descriptor.html#ab4c4694a381e4cfb44d304cf150743f9',1,'dlaf::common::DataDescriptor']]], - ['submatrixview_431',['SubMatrixView',['../structdlaf_1_1matrix_1_1_sub_matrix_view.html',1,'dlaf::matrix::SubMatrixView'],['../structdlaf_1_1matrix_1_1_sub_matrix_view.html#a7c7cde76a018f82ab23c749de90fe46d',1,'dlaf::matrix::SubMatrixView::SubMatrixView()']]], - ['subpanelview_432',['SubPanelView',['../structdlaf_1_1matrix_1_1_sub_panel_view.html',1,'dlaf::matrix::SubPanelView'],['../structdlaf_1_1matrix_1_1_sub_panel_view.html#a556b70975f5928ca81f2196d549be232',1,'dlaf::matrix::SubPanelView::SubPanelView()']]], - ['subpipelinetag_433',['SubPipelineTag',['../structdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4_1_1_sub_pipeline_tag.html',1,'dlaf::matrix::Matrix< const T, D >']]], - ['subtilereference_434',['subTileReference',['../classdlaf_1_1matrix_1_1_tile.html#a670161e49f6a18b5979e9938e4e7e867',1,'dlaf::matrix::Tile::subTileReference()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a4adf525073d06baaa47b06886272da41',1,'dlaf::matrix::Tile< const T, D >::subTileReference()']]], - ['subtilespec_435',['SubTileSpec',['../structdlaf_1_1matrix_1_1_sub_tile_spec.html',1,'dlaf::matrix']]], - ['sum_436',['sum',['../util__math_8h.html#afa826e9affbc0fc9e9dbe24d233a9ffb',1,'dlaf::util::size_t::sum()'],['../util__math_8h.html#ae1dbf257cc807e5bdba5a63087e3a2bc',1,'dlaf::util::ptrdiff_t::sum()']]], - ['sweepworker_437',['SweepWorker',['../classdlaf_1_1eigensolver_1_1internal_1_1_sweep_worker.html',1,'dlaf::eigensolver::internal']]], - ['sweepworkerdist_438',['SweepWorkerDist',['../classdlaf_1_1eigensolver_1_1internal_1_1_sweep_worker_dist.html',1,'dlaf::eigensolver::internal']]] + ['samesign_375',['sameSign',['../util__math_8h.html#a0922553e7cd5d09cc857bb33e8a1b91b',1,'dlaf::util']]], + ['scal_2eh_376',['scal.h',['../scal_8h.html',1,'']]], + ['scheduleallreduce_377',['scheduleAllReduce',['../kernels_2all__reduce_8h.html#a8cb64ca1859aef0b95cdcfa8a1f05297',1,'dlaf::comm']]], + ['scheduleallreduceinplace_378',['scheduleAllReduceInPlace',['../kernels_2all__reduce_8h.html#a689b3d06eea6e5d2c9899521e970ecb0',1,'dlaf::comm']]], + ['scheduleallsump2p_379',['scheduleAllSumP2P',['../p2p__allsum_8h.html#ac786ec776d72b61529c8a79c5baffff4',1,'dlaf::comm']]], + ['schedulerecvbcast_380',['scheduleRecvBcast',['../kernels_2broadcast_8h.html#a3645d0cfc3caad43c37c355896a7215f',1,'dlaf::comm']]], + ['schedulereducerecvinplace_381',['scheduleReduceRecvInPlace',['../kernels_2reduce_8h.html#a89af8ab199d6ba8902cc962122477f00',1,'dlaf::comm']]], + ['schedulereducesend_382',['scheduleReduceSend',['../kernels_2reduce_8h.html#a60899e3a349b0defe79b1561045a373a',1,'dlaf::comm']]], + ['schedulers_2eh_383',['schedulers.h',['../schedulers_8h.html',1,'']]], + ['schedulesendbcast_384',['scheduleSendBcast',['../kernels_2broadcast_8h.html#a98b583f24caaccdf37f3db1f3b46b457',1,'dlaf::comm']]], + ['scopedinitializer_385',['ScopedInitializer',['../structdlaf_1_1_scoped_initializer.html',1,'dlaf']]], + ['select_386',['select',['../matrix_8h.html#a1abc47c06dc1c46c9bdc2425090a02ab',1,'dlaf::matrix']]], + ['selectgeneric_387',['selectGeneric',['../matrix_8h.html#ac3737882aaaf23b1f901da6396202ba1',1,'dlaf::matrix::internal']]], + ['selectread_388',['selectRead',['../matrix_8h.html#a02db5399b6bc7e53a162f5939ed7ca20',1,'dlaf::matrix']]], + ['send_389',['send',['../sync_2broadcast_8h.html#a313a9dc3d212f70d7af273fa0c681aee',1,'dlaf::comm::sync::broadcast']]], + ['send_5fto_390',['send_to',['../basic_8h.html#adfd3205605f800afb2208146565a3bd4',1,'dlaf::comm::sync']]], + ['sendersinglevaluetypeimpl_391',['SenderSingleValueTypeImpl',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl.html',1,'dlaf::internal']]], + ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20rwtype_2c_20rtype_2c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5ftype_3a_3aread_20_3e_20_3e_20_3e_20_3e_392',['SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::read > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01pika_356df456d49829a554002eacb85cc5b7.html',1,'dlaf::internal']]], + ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20rwtype_2c_20rtype_2c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5ftype_3a_3areadwrite_20_3e_20_3e_20_3e_20_3e_393',['SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::readwrite > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01pika_3535bbe9d77284105a6158ec315ee38c.html',1,'dlaf::internal']]], + ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20std_3a_3areference_5fwrapper_3c_20t_20_3e_20_3e_20_3e_20_3e_394',['SenderSingleValueTypeImpl< TypeList< TypeList< std::reference_wrapper< T > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01std_1b569951bd408b620bb545799a6089d25.html',1,'dlaf::internal']]], + ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20t_20_3e_20_3e_20_3e_395',['SenderSingleValueTypeImpl< TypeList< TypeList< T > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01_t_01_4_01_4_01_4.html',1,'dlaf::internal']]], + ['set_396',['set',['../util__matrix_8h.html#af07478cd4a633a1422e160863d3d8510',1,'dlaf::matrix::util::set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f, const blas::Op op)'],['../util__matrix_8h.html#a1f796f1fd4dd705c90ac141d687843b2',1,'dlaf::matrix::util::set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f)']]], + ['set0_397',['set0',['../lapack_2tile_8h.html#ab094eeac6052422b50a298b42a947251',1,'dlaf::tile::set0()'],['../util__matrix_8h.html#a79688b3669b67c143114cf2eef3ba69b',1,'dlaf::matrix::util::set0(pika::execution::thread_priority priority, Panel< axis, T, D, storage > &panel)'],['../util__matrix_8h.html#a6cdf6afc18a6456ed68b3fd1dffa9c39',1,'dlaf::matrix::util::set0(pika::execution::thread_priority priority, LocalTileIndex begin, LocalTileSize sz, Matrix< T, D > &matrix)']]], + ['set_5frandom_398',['set_random',['../util__matrix_8h.html#a8e8c0d0551fbda5968d6c84eb3c70fc0',1,'dlaf::matrix::util']]], + ['set_5frandom_5fhermitian_399',['set_random_hermitian',['../util__matrix_8h.html#a3f7866203ad62c615d11ca18af4d4476',1,'dlaf::matrix::util']]], + ['set_5frandom_5fhermitian_5fpositive_5fdefinite_400',['set_random_hermitian_positive_definite',['../util__matrix_8h.html#ae2afbf7cdec8225e544b7b71aa52ee58',1,'dlaf::matrix::util']]], + ['set_5frandom_5fhermitian_5fwith_5foffset_401',['set_random_hermitian_with_offset',['../util__matrix_8h.html#a2f1823d0004b53105c4d4b7373208133',1,'dlaf::matrix::util::internal']]], + ['setheight_402',['setHeight',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a5ffd11dcc29ca2731c0cf232cb599d8e',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setrange_403',['setRange',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a88f387bed0a5cf8dabd5f6c5b3d2c3b0',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setrangeend_404',['setRangeEnd',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a542319871ee2379e84651d7dd91f0749',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setrangestart_405',['setRangeStart',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a621d6ffd26cb962792c4087159df2769',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['settile_406',['setTile',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa9c0ce49a7829488736d99011c5ac6c9',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setupinternalmatrix_407',['setupInternalMatrix',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a11f5026c35f76adb7c99c725dc969c44',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setwidth_408',['setWidth',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a74e5d5d87edfb1d2de8a54cfa812bd04',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['sharereadwritetile_409',['shareReadWriteTile',['../matrix_2tile_8h.html#aa8e4a64946c725cef06834fb80b75276',1,'dlaf::matrix']]], + ['single_5fthreaded_5fblas_2eh_410',['single_threaded_blas.h',['../single__threaded__blas_8h.html',1,'']]], + ['single_5ftile_5fper_5fblock_411',['single_tile_per_block',['../util__matrix_8h.html#a4e47364118c7bc9bb3d038ae4ec8a35c',1,'dlaf::matrix']]], + ['singlethreadedblasscope_412',['SingleThreadedBlasScope',['../classdlaf_1_1common_1_1internal_1_1_single_threaded_blas_scope.html',1,'dlaf::common::internal']]], + ['size_413',['size',['../classdlaf_1_1memory_1_1_memory_chunk.html#a886c87812f0a8eb9071c70477ac9f042',1,'dlaf::memory::MemoryChunk::size()'],['../classdlaf_1_1comm_1_1_communicator.html#a415f5d36c3033320074df329163c3127',1,'dlaf::comm::Communicator::size()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#a9307e8d79bffad58211c317c92c7dc77',1,'dlaf::comm::CommunicatorGrid::size()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#adc5a25970dab7cdea07e2f98328df805',1,'dlaf::matrix::internal::MatrixBase::size()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a7ef713ad7d24caccfc0863ab4b8481f2',1,'dlaf::matrix::Tile< const T, D >::size()'],['../classdlaf_1_1memory_1_1_memory_view.html#a5ae9491601035e4f2fd87352916f84c2',1,'dlaf::memory::MemoryView::size()']]], + ['size2d_414',['Size2D',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common::Size2D< IndexT, Tag >'],['../communicator__grid_8h.html#a1af44f2ea119a10f873e64cf048ba94a',1,'dlaf::comm::Size2D()']]], + ['size2d_3c_20indext_5fmpi_2c_20tag_5fmpi_20_3e_415',['Size2D< IndexT_MPI, TAG_MPI >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3aglobalelement_5ftag_20_3e_416',['Size2D< SizeType, matrix::GlobalElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3aglobaltile_5ftag_20_3e_417',['Size2D< SizeType, matrix::GlobalTile_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3alocalelement_5ftag_20_3e_418',['Size2D< SizeType, matrix::LocalElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3alocaltile_5ftag_20_3e_419',['Size2D< SizeType, matrix::LocalTile_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3atileelement_5ftag_20_3e_420',['Size2D< SizeType, matrix::TileElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['solver_2eh_421',['solver.h',['../solver_8h.html',1,'']]], + ['source_5flocation_422',['source_location',['../structdlaf_1_1common_1_1internal_1_1source__location.html',1,'dlaf::common::internal']]], + ['source_5flocation_423',['SOURCE_LOCATION',['../source__location_8h.html#aa6e6fc9684bd84ce7b72da6a2eef668f',1,'source_location.h']]], + ['source_5flocation_2eh_424',['source_location.h',['../source__location_8h.html',1,'']]], + ['splittile_425',['splitTile',['../matrix_2tile_8h.html#ae46de76ccc35985387f6820b1db9fb51',1,'dlaf::matrix::splitTile(ReadOnlyTileSender< T, D > tile, const SubTileSpec &spec)'],['../matrix_2tile_8h.html#a7a2383f364417fce38d6e7038cb18bd1',1,'dlaf::matrix::splitTile(ReadOnlyTileSender< T, D > tile, const std::vector< SubTileSpec > &specs)'],['../matrix_2tile_8h.html#ae5697bd5c424a95890978dfdc4739bf6',1,'dlaf::matrix::splitTile(ReadWriteTileSender< T, D > &&tile, const SubTileSpec &spec)']]], + ['splittiledisjoint_426',['splitTileDisjoint',['../matrix_2tile_8h.html#a4c15e2e57d6336347e6e19d945bd650e',1,'dlaf::matrix']]], + ['square_5fblocksize_427',['square_blocksize',['../util__matrix_8h.html#a355682eaf279129f2aa06c8e96eaa61b',1,'dlaf::matrix']]], + ['square_5fsize_428',['square_size',['../util__tile_8h.html#adfd1e0b37d590f9f6b56cdb276688f0f',1,'dlaf::tile::square_size()'],['../util__matrix_8h.html#aae6464887dae32e01668a32e71397cb0',1,'dlaf::matrix::square_size()']]], + ['start_5f_429',['start_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ac49583acc02b6ba394d9b4e3b7450e53',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['start_5foffset_5f_430',['start_offset_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a8603ffce77527952a466a9630d98cd5c',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['stedc_431',['stedc',['../lapack_2tile_8h.html#a68163899b7b15f447fdbb05e475d9cc6',1,'dlaf::tile']]], + ['stride_432',['stride',['../structdlaf_1_1common_1_1_data_descriptor.html#ab4c4694a381e4cfb44d304cf150743f9',1,'dlaf::common::DataDescriptor']]], + ['submatrixview_433',['SubMatrixView',['../structdlaf_1_1matrix_1_1_sub_matrix_view.html',1,'dlaf::matrix::SubMatrixView'],['../structdlaf_1_1matrix_1_1_sub_matrix_view.html#a7c7cde76a018f82ab23c749de90fe46d',1,'dlaf::matrix::SubMatrixView::SubMatrixView()']]], + ['subpanelview_434',['SubPanelView',['../structdlaf_1_1matrix_1_1_sub_panel_view.html',1,'dlaf::matrix::SubPanelView'],['../structdlaf_1_1matrix_1_1_sub_panel_view.html#a556b70975f5928ca81f2196d549be232',1,'dlaf::matrix::SubPanelView::SubPanelView()']]], + ['subpipeline_435',['subPipeline',['../classdlaf_1_1matrix_1_1_matrix.html#a720bcfeb6db4c2f2d6e144f8570b8e23',1,'dlaf::matrix::Matrix']]], + ['subpipelineconst_436',['subPipelineConst',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#aa00bb4dd8e50ec14ad04c8d600982d8c',1,'dlaf::matrix::Matrix< const T, D >']]], + ['subpipelinetag_437',['SubPipelineTag',['../structdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4_1_1_sub_pipeline_tag.html',1,'dlaf::matrix::Matrix< const T, D >']]], + ['subtilereference_438',['subTileReference',['../classdlaf_1_1matrix_1_1_tile.html#a670161e49f6a18b5979e9938e4e7e867',1,'dlaf::matrix::Tile::subTileReference()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a4adf525073d06baaa47b06886272da41',1,'dlaf::matrix::Tile< const T, D >::subTileReference()']]], + ['subtilespec_439',['SubTileSpec',['../structdlaf_1_1matrix_1_1_sub_tile_spec.html',1,'dlaf::matrix']]], + ['sum_440',['sum',['../util__math_8h.html#afa826e9affbc0fc9e9dbe24d233a9ffb',1,'dlaf::util::size_t::sum()'],['../util__math_8h.html#ae1dbf257cc807e5bdba5a63087e3a2bc',1,'dlaf::util::ptrdiff_t::sum()']]], + ['sweepworker_441',['SweepWorker',['../classdlaf_1_1eigensolver_1_1internal_1_1_sweep_worker.html',1,'dlaf::eigensolver::internal']]], + ['sweepworkerdist_442',['SweepWorkerDist',['../classdlaf_1_1eigensolver_1_1internal_1_1_sweep_worker_dist.html',1,'dlaf::eigensolver::internal']]] ]; diff --git a/master/search/all_12.js b/master/search/all_12.js index 2422ee6c35..cf59bed300 100644 --- a/master/search/all_12.js +++ b/master/search/all_12.js @@ -1,55 +1,55 @@ var searchData= [ - ['tile_439',['Tile',['../classdlaf_1_1matrix_1_1_tile.html',1,'dlaf::matrix::Tile< T, D >'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a38b4f570cc1ccedb8880d469d353414e',1,'dlaf::matrix::Tile< const T, D >::Tile()=default'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a73f2bfb1856011bce752fca6582ba584',1,'dlaf::matrix::Tile< const T, D >::Tile(const TileElementSize &size, memory::MemoryView< ElementType, D > &&memory_view, SizeType ld) noexcept'],['../classdlaf_1_1matrix_1_1_tile.html#af4bddf624138747a5089071f1f57eee4',1,'dlaf::matrix::Tile::Tile()=default'],['../classdlaf_1_1matrix_1_1_tile.html#a4c7d7a75afcd2cc6911b40643f4f095f',1,'dlaf::matrix::Tile::Tile(const TileElementSize &size, memory::MemoryView< ElementType, D > &&memory_view, SizeType ld) noexcept']]], - ['tile_2eh_440',['tile.h',['../blas_2tile_8h.html',1,'(Global Namespace)'],['../lapack_2tile_8h.html',1,'(Global Namespace)'],['../matrix_2tile_8h.html',1,'(Global Namespace)']]], - ['tile_3c_20const_20t_2c_20d_20_3e_441',['Tile< const T, D >',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html',1,'dlaf::matrix']]], - ['tile_5fextensions_2eh_442',['tile_extensions.h',['../tile__extensions_8h.html',1,'']]], - ['tileaccesshelper_443',['TileAccessHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_tile_access_helper.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['tilecollector_444',['TileCollector',['../classdlaf_1_1eigensolver_1_1internal_1_1_tile_collector.html',1,'dlaf::eigensolver::internal']]], - ['tiledata_445',['TileData',['../classdlaf_1_1matrix_1_1internal_1_1_tile_data.html',1,'dlaf::matrix::internal']]], - ['tileelementfromelement_446',['tileElementFromElement',['../util__distribution_8h.html#a77c1788e03f30e7d9511a7e90a89a494',1,'dlaf::util::matrix']]], - ['tileelementfromglobalelement_447',['tileElementFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#afe2b66a02c3a1a78bfac1bb2f20f2864',1,'dlaf::matrix::Distribution']]], - ['tileelementindex_448',['tileElementIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a71ad776a93750b9c6d1ebfa2d36e28db',1,'dlaf::matrix::Distribution']]], - ['tilefromelement_449',['tileFromElement',['../util__distribution_8h.html#a8876d879740e8f24112ca972961e11c7',1,'dlaf::util::matrix']]], - ['tilelayout_450',['tileLayout',['../layout__info_8h.html#a94ba1a30c8eef4798f864ae119c77113',1,'dlaf::matrix::tileLayout(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col)'],['../layout__info_8h.html#a970afb0f1a7abc87e507e6c4d35f442a',1,'dlaf::matrix::tileLayout(const matrix::Distribution &distribution, SizeType ld_tile, SizeType tiles_per_col)'],['../layout__info_8h.html#a254e1e0948e647642b0f2615aeaa0ea2',1,'dlaf::matrix::tileLayout(const LocalElementSize &size, const TileElementSize &block_size)'],['../layout__info_8h.html#a873adaaaa2e661caed98af371fbb1943',1,'dlaf::matrix::tileLayout(const matrix::Distribution &distribution)']]], - ['tilelinearindex_451',['tileLinearIndex',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a5cdaa0376724e6f624591a6a6bcf46d9',1,'dlaf::matrix::internal::MatrixBase']]], - ['tileoffset_452',['tileOffset',['../classdlaf_1_1matrix_1_1_layout_info.html#ad03e1291efd2ed416ace133a24a30ac4',1,'dlaf::matrix::LayoutInfo']]], - ['tilepipeline_453',['TilePipeline',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html',1,'dlaf::matrix::internal']]], - ['tilesize_454',['tileSize',['../classdlaf_1_1matrix_1_1_layout_info.html#a62e3619b9016e9c827c120fce629d08e',1,'dlaf::matrix::LayoutInfo::tileSize()'],['../classdlaf_1_1matrix_1_1_distribution.html#a4cec137b1316902f68b63148d282be62',1,'dlaf::matrix::Distribution::tileSize()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a891c1da69cefcfdfb73b3915ccc2cce5',1,'dlaf::matrix::internal::MatrixBase::tileSize()']]], - ['tilesizefromglobalelement_455',['tileSizeFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a3dd77d459960a11b56fc1e7fb27932f7',1,'dlaf::matrix::Distribution']]], - ['timer_456',['Timer',['../classdlaf_1_1common_1_1_timer.html',1,'dlaf::common']]], - ['timer_2eh_457',['timer.h',['../timer_8h.html',1,'']]], - ['to_5fint_458',['to_int',['../types_8h.html#af90e7fd5acadf5987b7199b0bd44deea',1,'dlaf']]], - ['to_5fsigned_459',['to_signed',['../types_8h.html#a2246feaf3c49f3f3ae5928082363f7d3',1,'dlaf::to_signed(const SB value)'],['../types_8h.html#aa39d2d6e54b007c18e1c57ff380c73cd',1,'dlaf::to_signed(const U unsigned_value)']]], - ['to_5fsizet_460',['to_sizet',['../types_8h.html#a25d16ff6621f011d9a98be3d9ebf29f1',1,'dlaf']]], - ['to_5fsizetype_461',['to_SizeType',['../types_8h.html#ab1f0edc8e1281293c9c1325cb9d4bec9',1,'dlaf']]], - ['to_5fuint_462',['to_uint',['../types_8h.html#a4543febf91f08c23fd8a6eeca8f49d82',1,'dlaf']]], - ['to_5funsigned_463',['to_unsigned',['../types_8h.html#ae9973646cf4ce493ed1831a5bdbf7500',1,'dlaf::to_unsigned(const UB unsigned_value)'],['../types_8h.html#afeb5db746331f9ca3a669373e1af84ba',1,'dlaf::to_unsigned(const S signed_value)']]], - ['total_5fops_464',['total_ops',['../types_8h.html#a831d9cfa2a32641121185df5d9d70cdc',1,'dlaf']]], - ['traits_2eh_465',['traits.h',['../traits_8h.html',1,'']]], - ['transformpack_466',['TransformPack',['../structdlaf_1_1internal_1_1_transform_pack.html',1,'dlaf::internal']]], - ['transformpack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20transformer_20_3e_467',['TransformPack< Pack< Ts... >, Transformer >',['../structdlaf_1_1internal_1_1_transform_pack_3_01_pack_3_01_ts_8_8_8_01_4_00_01_transformer_01_4.html',1,'dlaf::internal']]], - ['transpose_468',['transpose',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a98f8b3b0cd7b4c1fa62b40b9de0041bb',1,'dlaf::common::internal::basic_coords']]], - ['transposed_469',['transposed',['../index2d_8h.html#a024753ac73fbefa174548c575327862b',1,'dlaf::common']]], - ['triangular_470',['Triangular',['../structdlaf_1_1multiplication_1_1internal_1_1_triangular.html',1,'dlaf::multiplication::internal::Triangular< backend, device, T >'],['../structdlaf_1_1solver_1_1internal_1_1_triangular.html',1,'dlaf::solver::internal::Triangular< backend, device, T >']]], - ['triangular_471',['triangular',['../multiplication_2triangular_8h.html#a6300b8eb1a939512a5b1c45286e89f51',1,'dlaf::multiplication::triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../multiplication_2triangular_8h.html#ad4de985e76a6c12c487b36fa773b870b',1,'dlaf::multiplication::triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../solver_2triangular_8h.html#a9d48ec5978acfd3714a66613678a1afd',1,'dlaf::solver::triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../solver_2triangular_8h.html#a3ff8022461abe72c4197e34601d592a9',1,'dlaf::solver::triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)']]], - ['triangular_2eh_472',['triangular.h',['../solver_2triangular_8h.html',1,'(Global Namespace)'],['../multiplication_2triangular_8h.html',1,'(Global Namespace)']]], - ['tridiag_5fsolver_2eh_473',['tridiag_solver.h',['../tridiag__solver_8h.html',1,'']]], - ['tridiagresult_474',['TridiagResult',['../structdlaf_1_1eigensolver_1_1_tridiag_result.html',1,'dlaf::eigensolver']]], - ['tridiagsolver_475',['TridiagSolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_tridiag_solver.html',1,'dlaf::eigensolver::internal']]], - ['tridiagsolver_476',['tridiagSolver',['../tridiag__solver_8h.html#a52f0bd4b681a2d1c3e47bfd6a6ebf15a',1,'dlaf::eigensolver::tridiagSolver(Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, device > &evals, Matrix< T, device > &evecs)'],['../tridiag__solver_8h.html#a0cfeab14e9e1dc7a2fb703006cd7e957',1,'dlaf::eigensolver::tridiagSolver(comm::CommunicatorGrid grid, Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, D > &evals, Matrix< T, D > &evecs)']]], - ['trmm_477',['trmm',['../blas_2tile_8h.html#a61397e6748a24f13ee6ad7f23e5339f7',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< T, D > &b)'],['../blas_2tile_8h.html#a185aa2ac627de6e9e393240cda2b5f4c',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a383ebe00fbfd1f328dd0378b8aa42b89',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &p)']]], - ['trmm3_478',['trmm3',['../blas_2tile_8h.html#aebff48d52916dafd14479dc3b1e04e8e',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const Tile< T, D > &c)'],['../blas_2tile_8h.html#a9906093b08eefa0a5341b140ce6100a3',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#a0cdf7b22f8a310978252ff7a0c456b98',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &p, Sender &&s)']]], - ['trmmsizes_479',['trmmSizes',['../structdlaf_1_1tile_1_1internal_1_1trmm_sizes.html',1,'dlaf::tile::internal']]], - ['trsm_480',['trsm',['../blas_2tile_8h.html#a2fe48f2d3a3a7515a99805d135c5b2ce',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< T, D > &b)'],['../blas_2tile_8h.html#a1b11d671c920bf9dce76092e6d1aaa8c',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a23eb607d942637d4e14331a2f0f253f7',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &p)']]], - ['trsmsizes_481',['trsmSizes',['../structdlaf_1_1tile_1_1internal_1_1trsm_sizes.html',1,'dlaf::tile::internal']]], - ['tune_2eh_482',['tune.h',['../tune_8h.html',1,'']]], - ['tuneparameters_483',['TuneParameters',['../structdlaf_1_1_tune_parameters.html',1,'dlaf']]], - ['type_5fhandler_484',['type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html',1,'dlaf::comm::internal::type_handler< T >'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#aa9eebfb39b8aabe2344f61a7bf26ee40',1,'dlaf::comm::internal::type_handler::type_handler(SizeType nblocks, SizeType block_size, SizeType stride)'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#aa71d40004e1c645161b1d00b511ad9ee',1,'dlaf::comm::internal::type_handler::type_handler() noexcept=default']]], - ['type_5fhandler_2eh_485',['type_handler.h',['../type__handler_8h.html',1,'']]], - ['typeinfo_486',['TypeInfo',['../structdlaf_1_1_type_info.html',1,'dlaf']]], - ['typeinfo_3c_20const_20t_20_3e_487',['TypeInfo< const T >',['../structdlaf_1_1_type_info_3_01const_01_t_01_4.html',1,'dlaf']]], - ['typeinfo_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_488',['TypeInfo< std::complex< T > >',['../structdlaf_1_1_type_info_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf']]], - ['typelist_489',['TypeList',['../structdlaf_1_1internal_1_1_type_list.html',1,'dlaf::internal']]], - ['types_2eh_490',['types.h',['../types_8h.html',1,'']]] + ['tile_443',['Tile',['../classdlaf_1_1matrix_1_1_tile.html',1,'dlaf::matrix::Tile< T, D >'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a38b4f570cc1ccedb8880d469d353414e',1,'dlaf::matrix::Tile< const T, D >::Tile()=default'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a73f2bfb1856011bce752fca6582ba584',1,'dlaf::matrix::Tile< const T, D >::Tile(const TileElementSize &size, memory::MemoryView< ElementType, D > &&memory_view, SizeType ld) noexcept'],['../classdlaf_1_1matrix_1_1_tile.html#af4bddf624138747a5089071f1f57eee4',1,'dlaf::matrix::Tile::Tile()=default'],['../classdlaf_1_1matrix_1_1_tile.html#a4c7d7a75afcd2cc6911b40643f4f095f',1,'dlaf::matrix::Tile::Tile(const TileElementSize &size, memory::MemoryView< ElementType, D > &&memory_view, SizeType ld) noexcept']]], + ['tile_2eh_444',['tile.h',['../blas_2tile_8h.html',1,'(Global Namespace)'],['../lapack_2tile_8h.html',1,'(Global Namespace)'],['../matrix_2tile_8h.html',1,'(Global Namespace)']]], + ['tile_3c_20const_20t_2c_20d_20_3e_445',['Tile< const T, D >',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html',1,'dlaf::matrix']]], + ['tile_5fextensions_2eh_446',['tile_extensions.h',['../tile__extensions_8h.html',1,'']]], + ['tileaccesshelper_447',['TileAccessHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_tile_access_helper.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['tilecollector_448',['TileCollector',['../classdlaf_1_1eigensolver_1_1internal_1_1_tile_collector.html',1,'dlaf::eigensolver::internal']]], + ['tiledata_449',['TileData',['../classdlaf_1_1matrix_1_1internal_1_1_tile_data.html',1,'dlaf::matrix::internal']]], + ['tileelementfromelement_450',['tileElementFromElement',['../util__distribution_8h.html#a77c1788e03f30e7d9511a7e90a89a494',1,'dlaf::util::matrix']]], + ['tileelementfromglobalelement_451',['tileElementFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#afe2b66a02c3a1a78bfac1bb2f20f2864',1,'dlaf::matrix::Distribution']]], + ['tileelementindex_452',['tileElementIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a71ad776a93750b9c6d1ebfa2d36e28db',1,'dlaf::matrix::Distribution']]], + ['tilefromelement_453',['tileFromElement',['../util__distribution_8h.html#a8876d879740e8f24112ca972961e11c7',1,'dlaf::util::matrix']]], + ['tilelayout_454',['tileLayout',['../layout__info_8h.html#a94ba1a30c8eef4798f864ae119c77113',1,'dlaf::matrix::tileLayout(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col)'],['../layout__info_8h.html#a970afb0f1a7abc87e507e6c4d35f442a',1,'dlaf::matrix::tileLayout(const matrix::Distribution &distribution, SizeType ld_tile, SizeType tiles_per_col)'],['../layout__info_8h.html#a254e1e0948e647642b0f2615aeaa0ea2',1,'dlaf::matrix::tileLayout(const LocalElementSize &size, const TileElementSize &block_size)'],['../layout__info_8h.html#a873adaaaa2e661caed98af371fbb1943',1,'dlaf::matrix::tileLayout(const matrix::Distribution &distribution)']]], + ['tilelinearindex_455',['tileLinearIndex',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a5cdaa0376724e6f624591a6a6bcf46d9',1,'dlaf::matrix::internal::MatrixBase']]], + ['tileoffset_456',['tileOffset',['../classdlaf_1_1matrix_1_1_layout_info.html#ad03e1291efd2ed416ace133a24a30ac4',1,'dlaf::matrix::LayoutInfo']]], + ['tilepipeline_457',['TilePipeline',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html',1,'dlaf::matrix::internal']]], + ['tilesize_458',['tileSize',['../classdlaf_1_1matrix_1_1_layout_info.html#a62e3619b9016e9c827c120fce629d08e',1,'dlaf::matrix::LayoutInfo::tileSize()'],['../classdlaf_1_1matrix_1_1_distribution.html#a4cec137b1316902f68b63148d282be62',1,'dlaf::matrix::Distribution::tileSize()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a891c1da69cefcfdfb73b3915ccc2cce5',1,'dlaf::matrix::internal::MatrixBase::tileSize()']]], + ['tilesizefromglobalelement_459',['tileSizeFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a3dd77d459960a11b56fc1e7fb27932f7',1,'dlaf::matrix::Distribution']]], + ['timer_460',['Timer',['../classdlaf_1_1common_1_1_timer.html',1,'dlaf::common']]], + ['timer_2eh_461',['timer.h',['../timer_8h.html',1,'']]], + ['to_5fint_462',['to_int',['../types_8h.html#af90e7fd5acadf5987b7199b0bd44deea',1,'dlaf']]], + ['to_5fsigned_463',['to_signed',['../types_8h.html#a2246feaf3c49f3f3ae5928082363f7d3',1,'dlaf::to_signed(const SB value)'],['../types_8h.html#aa39d2d6e54b007c18e1c57ff380c73cd',1,'dlaf::to_signed(const U unsigned_value)']]], + ['to_5fsizet_464',['to_sizet',['../types_8h.html#a25d16ff6621f011d9a98be3d9ebf29f1',1,'dlaf']]], + ['to_5fsizetype_465',['to_SizeType',['../types_8h.html#ab1f0edc8e1281293c9c1325cb9d4bec9',1,'dlaf']]], + ['to_5fuint_466',['to_uint',['../types_8h.html#a4543febf91f08c23fd8a6eeca8f49d82',1,'dlaf']]], + ['to_5funsigned_467',['to_unsigned',['../types_8h.html#ae9973646cf4ce493ed1831a5bdbf7500',1,'dlaf::to_unsigned(const UB unsigned_value)'],['../types_8h.html#afeb5db746331f9ca3a669373e1af84ba',1,'dlaf::to_unsigned(const S signed_value)']]], + ['total_5fops_468',['total_ops',['../types_8h.html#a831d9cfa2a32641121185df5d9d70cdc',1,'dlaf']]], + ['traits_2eh_469',['traits.h',['../traits_8h.html',1,'']]], + ['transformpack_470',['TransformPack',['../structdlaf_1_1internal_1_1_transform_pack.html',1,'dlaf::internal']]], + ['transformpack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20transformer_20_3e_471',['TransformPack< Pack< Ts... >, Transformer >',['../structdlaf_1_1internal_1_1_transform_pack_3_01_pack_3_01_ts_8_8_8_01_4_00_01_transformer_01_4.html',1,'dlaf::internal']]], + ['transpose_472',['transpose',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a98f8b3b0cd7b4c1fa62b40b9de0041bb',1,'dlaf::common::internal::basic_coords']]], + ['transposed_473',['transposed',['../index2d_8h.html#a024753ac73fbefa174548c575327862b',1,'dlaf::common']]], + ['triangular_474',['Triangular',['../structdlaf_1_1multiplication_1_1internal_1_1_triangular.html',1,'dlaf::multiplication::internal::Triangular< backend, device, T >'],['../structdlaf_1_1solver_1_1internal_1_1_triangular.html',1,'dlaf::solver::internal::Triangular< backend, device, T >']]], + ['triangular_475',['triangular',['../multiplication_2triangular_8h.html#a6300b8eb1a939512a5b1c45286e89f51',1,'dlaf::multiplication::triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../multiplication_2triangular_8h.html#ad4de985e76a6c12c487b36fa773b870b',1,'dlaf::multiplication::triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../solver_2triangular_8h.html#a9d48ec5978acfd3714a66613678a1afd',1,'dlaf::solver::triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../solver_2triangular_8h.html#a3ff8022461abe72c4197e34601d592a9',1,'dlaf::solver::triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)']]], + ['triangular_2eh_476',['triangular.h',['../solver_2triangular_8h.html',1,'(Global Namespace)'],['../multiplication_2triangular_8h.html',1,'(Global Namespace)']]], + ['tridiag_5fsolver_2eh_477',['tridiag_solver.h',['../tridiag__solver_8h.html',1,'']]], + ['tridiagresult_478',['TridiagResult',['../structdlaf_1_1eigensolver_1_1_tridiag_result.html',1,'dlaf::eigensolver']]], + ['tridiagsolver_479',['TridiagSolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_tridiag_solver.html',1,'dlaf::eigensolver::internal']]], + ['tridiagsolver_480',['tridiagSolver',['../tridiag__solver_8h.html#a52f0bd4b681a2d1c3e47bfd6a6ebf15a',1,'dlaf::eigensolver::tridiagSolver(Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, device > &evals, Matrix< T, device > &evecs)'],['../tridiag__solver_8h.html#a0cfeab14e9e1dc7a2fb703006cd7e957',1,'dlaf::eigensolver::tridiagSolver(comm::CommunicatorGrid grid, Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, D > &evals, Matrix< T, D > &evecs)']]], + ['trmm_481',['trmm',['../blas_2tile_8h.html#a61397e6748a24f13ee6ad7f23e5339f7',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< T, D > &b)'],['../blas_2tile_8h.html#a185aa2ac627de6e9e393240cda2b5f4c',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a383ebe00fbfd1f328dd0378b8aa42b89',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &p)']]], + ['trmm3_482',['trmm3',['../blas_2tile_8h.html#aebff48d52916dafd14479dc3b1e04e8e',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const Tile< T, D > &c)'],['../blas_2tile_8h.html#a9906093b08eefa0a5341b140ce6100a3',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#a0cdf7b22f8a310978252ff7a0c456b98',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &p, Sender &&s)']]], + ['trmmsizes_483',['trmmSizes',['../structdlaf_1_1tile_1_1internal_1_1trmm_sizes.html',1,'dlaf::tile::internal']]], + ['trsm_484',['trsm',['../blas_2tile_8h.html#a2fe48f2d3a3a7515a99805d135c5b2ce',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< T, D > &b)'],['../blas_2tile_8h.html#a1b11d671c920bf9dce76092e6d1aaa8c',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a23eb607d942637d4e14331a2f0f253f7',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &p)']]], + ['trsmsizes_485',['trsmSizes',['../structdlaf_1_1tile_1_1internal_1_1trsm_sizes.html',1,'dlaf::tile::internal']]], + ['tune_2eh_486',['tune.h',['../tune_8h.html',1,'']]], + ['tuneparameters_487',['TuneParameters',['../structdlaf_1_1_tune_parameters.html',1,'dlaf']]], + ['type_5fhandler_488',['type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html',1,'dlaf::comm::internal::type_handler< T >'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#aa9eebfb39b8aabe2344f61a7bf26ee40',1,'dlaf::comm::internal::type_handler::type_handler(SizeType nblocks, SizeType block_size, SizeType stride)'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#aa71d40004e1c645161b1d00b511ad9ee',1,'dlaf::comm::internal::type_handler::type_handler() noexcept=default']]], + ['type_5fhandler_2eh_489',['type_handler.h',['../type__handler_8h.html',1,'']]], + ['typeinfo_490',['TypeInfo',['../structdlaf_1_1_type_info.html',1,'dlaf']]], + ['typeinfo_3c_20const_20t_20_3e_491',['TypeInfo< const T >',['../structdlaf_1_1_type_info_3_01const_01_t_01_4.html',1,'dlaf']]], + ['typeinfo_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_492',['TypeInfo< std::complex< T > >',['../structdlaf_1_1_type_info_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf']]], + ['typelist_493',['TypeList',['../structdlaf_1_1internal_1_1_type_list.html',1,'dlaf::internal']]], + ['types_2eh_494',['types.h',['../types_8h.html',1,'']]] ]; diff --git a/master/search/all_13.js b/master/search/all_13.js index 152dc03654..65b5c7a8ef 100644 --- a/master/search/all_13.js +++ b/master/search/all_13.js @@ -1,27 +1,27 @@ var searchData= [ - ['unique_491',['Unique',['../structdlaf_1_1internal_1_1_unique.html',1,'dlaf::internal']]], - ['unique_3c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_492',['Unique< Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_3_01_pack_3_01_ts_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], - ['uniquehelper_493',['UniqueHelper',['../structdlaf_1_1internal_1_1_unique_helper.html',1,'dlaf::internal']]], - ['uniquehelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_20u_2c_20us_2e_2e_2e_20_3e_20_3e_494',['UniqueHelper< Pack< Ts... >, Pack< U, Us... > >',['../structdlaf_1_1internal_1_1_unique_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_01_u_00_01_us_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], - ['uniquehelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_3e_20_3e_495',['UniqueHelper< Pack< Ts... >, Pack<> >',['../structdlaf_1_1internal_1_1_unique_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_4_01_4.html',1,'dlaf::internal']]], - ['uniquehelper_3c_20pack_3c_3e_2c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_496',['UniqueHelper< Pack<>, Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_helper.html',1,'dlaf::internal']]], - ['uniquepack_497',['UniquePack',['../structdlaf_1_1internal_1_1_unique_pack.html',1,'dlaf::internal']]], - ['uniquepack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_498',['UniquePack< Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_pack_3_01_pack_3_01_ts_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], - ['uniquepackhelper_499',['UniquePackHelper',['../structdlaf_1_1internal_1_1_unique_pack_helper.html',1,'dlaf::internal']]], - ['uniquepackhelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_20u_2c_20us_2e_2e_2e_20_3e_20_3e_500',['UniquePackHelper< Pack< Ts... >, Pack< U, Us... > >',['../structdlaf_1_1internal_1_1_unique_pack_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_01_u_00_01_us_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], - ['uniquepackhelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_3e_20_3e_501',['UniquePackHelper< Pack< Ts... >, Pack<> >',['../structdlaf_1_1internal_1_1_unique_pack_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_4_01_4.html',1,'dlaf::internal']]], - ['uniquepackhelper_3c_20pack_3c_3e_2c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_502',['UniquePackHelper< Pack<>, Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_pack_helper.html',1,'dlaf::internal']]], - ['unwrap_503',['unwrap',['../unwrap_8h.html#a8da70a5535ac987a8f6ca7f8b195547d',1,'dlaf::common::internal']]], - ['unwrap_2eh_504',['unwrap.h',['../unwrap_8h.html',1,'']]], - ['unwrapper_505',['Unwrapper',['../structdlaf_1_1common_1_1internal_1_1_unwrapper.html',1,'dlaf::common::internal']]], - ['unwrapper_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20t1_2c_20t2_2c_20at_20_3e_20_3e_506',['Unwrapper< pika::execution::experimental::async_rw_mutex_access_wrapper< T1, T2, at > >',['../structdlaf_1_1common_1_1internal_1_1_unwrapper_3_01pika_1_1execution_1_1experimental_1_1async__r5be7b0afcb71e6df73ebcc954d80821d.html',1,'dlaf::common::internal']]], - ['unwrapper_3c_20std_3a_3areference_5fwrapper_3c_20t_20_3e_20_3e_507',['Unwrapper< std::reference_wrapper< T > >',['../structdlaf_1_1common_1_1internal_1_1_unwrapper_3_01std_1_1reference__wrapper_3_01_t_01_4_01_4.html',1,'dlaf::common::internal']]], - ['unwrapping_508',['Unwrapping',['../structdlaf_1_1common_1_1internal_1_1_unwrapping.html',1,'dlaf::common::internal']]], - ['util_5fblas_2eh_509',['util_blas.h',['../util__blas_8h.html',1,'']]], - ['util_5fdistribution_2eh_510',['util_distribution.h',['../util__distribution_8h.html',1,'']]], - ['util_5flapack_2eh_511',['util_lapack.h',['../util__lapack_8h.html',1,'']]], - ['util_5fmath_2eh_512',['util_math.h',['../util__math_8h.html',1,'']]], - ['util_5fmatrix_2eh_513',['util_matrix.h',['../util__matrix_8h.html',1,'']]], - ['util_5ftile_2eh_514',['util_tile.h',['../util__tile_8h.html',1,'']]] + ['unique_495',['Unique',['../structdlaf_1_1internal_1_1_unique.html',1,'dlaf::internal']]], + ['unique_3c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_496',['Unique< Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_3_01_pack_3_01_ts_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], + ['uniquehelper_497',['UniqueHelper',['../structdlaf_1_1internal_1_1_unique_helper.html',1,'dlaf::internal']]], + ['uniquehelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_20u_2c_20us_2e_2e_2e_20_3e_20_3e_498',['UniqueHelper< Pack< Ts... >, Pack< U, Us... > >',['../structdlaf_1_1internal_1_1_unique_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_01_u_00_01_us_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], + ['uniquehelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_3e_20_3e_499',['UniqueHelper< Pack< Ts... >, Pack<> >',['../structdlaf_1_1internal_1_1_unique_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_4_01_4.html',1,'dlaf::internal']]], + ['uniquehelper_3c_20pack_3c_3e_2c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_500',['UniqueHelper< Pack<>, Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_helper.html',1,'dlaf::internal']]], + ['uniquepack_501',['UniquePack',['../structdlaf_1_1internal_1_1_unique_pack.html',1,'dlaf::internal']]], + ['uniquepack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_502',['UniquePack< Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_pack_3_01_pack_3_01_ts_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], + ['uniquepackhelper_503',['UniquePackHelper',['../structdlaf_1_1internal_1_1_unique_pack_helper.html',1,'dlaf::internal']]], + ['uniquepackhelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_20u_2c_20us_2e_2e_2e_20_3e_20_3e_504',['UniquePackHelper< Pack< Ts... >, Pack< U, Us... > >',['../structdlaf_1_1internal_1_1_unique_pack_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_01_u_00_01_us_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], + ['uniquepackhelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_3e_20_3e_505',['UniquePackHelper< Pack< Ts... >, Pack<> >',['../structdlaf_1_1internal_1_1_unique_pack_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_4_01_4.html',1,'dlaf::internal']]], + ['uniquepackhelper_3c_20pack_3c_3e_2c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_506',['UniquePackHelper< Pack<>, Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_pack_helper.html',1,'dlaf::internal']]], + ['unwrap_507',['unwrap',['../unwrap_8h.html#a8da70a5535ac987a8f6ca7f8b195547d',1,'dlaf::common::internal']]], + ['unwrap_2eh_508',['unwrap.h',['../unwrap_8h.html',1,'']]], + ['unwrapper_509',['Unwrapper',['../structdlaf_1_1common_1_1internal_1_1_unwrapper.html',1,'dlaf::common::internal']]], + ['unwrapper_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20t1_2c_20t2_2c_20at_20_3e_20_3e_510',['Unwrapper< pika::execution::experimental::async_rw_mutex_access_wrapper< T1, T2, at > >',['../structdlaf_1_1common_1_1internal_1_1_unwrapper_3_01pika_1_1execution_1_1experimental_1_1async__r5be7b0afcb71e6df73ebcc954d80821d.html',1,'dlaf::common::internal']]], + ['unwrapper_3c_20std_3a_3areference_5fwrapper_3c_20t_20_3e_20_3e_511',['Unwrapper< std::reference_wrapper< T > >',['../structdlaf_1_1common_1_1internal_1_1_unwrapper_3_01std_1_1reference__wrapper_3_01_t_01_4_01_4.html',1,'dlaf::common::internal']]], + ['unwrapping_512',['Unwrapping',['../structdlaf_1_1common_1_1internal_1_1_unwrapping.html',1,'dlaf::common::internal']]], + ['util_5fblas_2eh_513',['util_blas.h',['../util__blas_8h.html',1,'']]], + ['util_5fdistribution_2eh_514',['util_distribution.h',['../util__distribution_8h.html',1,'']]], + ['util_5flapack_2eh_515',['util_lapack.h',['../util__lapack_8h.html',1,'']]], + ['util_5fmath_2eh_516',['util_math.h',['../util__math_8h.html',1,'']]], + ['util_5fmatrix_2eh_517',['util_matrix.h',['../util__matrix_8h.html',1,'']]], + ['util_5ftile_2eh_518',['util_tile.h',['../util__tile_8h.html',1,'']]] ]; diff --git a/master/search/all_14.js b/master/search/all_14.js index f3fc61255a..e55e969d87 100644 --- a/master/search/all_14.js +++ b/master/search/all_14.js @@ -1,10 +1,10 @@ var searchData= [ - ['vaccesshelper_515',['VAccessHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1_v_access_helper.html',1,'dlaf::eigensolver::internal']]], - ['valid_516',['valid',['../classdlaf_1_1common_1_1_pipeline.html#affda528ea4c5e0e54f8fca4250b53af5',1,'dlaf::common::Pipeline::valid()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a3e3096bb1ef85b6017bd8c42c9617586',1,'dlaf::matrix::internal::TilePipeline::valid()']]], - ['vector_517',['vector',['../structdlaf_1_1common_1_1internal_1_1vector.html',1,'dlaf::common::internal']]], - ['vector_2eh_518',['vector.h',['../vector_8h.html',1,'']]], - ['vector_3c_20readonlysendertype_20_3e_519',['vector< ReadOnlySenderType >',['../structdlaf_1_1common_1_1internal_1_1vector.html',1,'dlaf::common::internal']]], - ['view_520',['View',['../structdlaf_1_1matrix_1_1internal_1_1_view.html',1,'dlaf::matrix::internal::View'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a3635af3164815f905a9fd33979a14a34',1,'dlaf::matrix::internal::View::View()']]], - ['views_2eh_521',['views.h',['../views_8h.html',1,'']]] + ['vaccesshelper_519',['VAccessHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1_v_access_helper.html',1,'dlaf::eigensolver::internal']]], + ['valid_520',['valid',['../classdlaf_1_1common_1_1_pipeline.html#affda528ea4c5e0e54f8fca4250b53af5',1,'dlaf::common::Pipeline::valid()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a3e3096bb1ef85b6017bd8c42c9617586',1,'dlaf::matrix::internal::TilePipeline::valid()']]], + ['vector_521',['vector',['../structdlaf_1_1common_1_1internal_1_1vector.html',1,'dlaf::common::internal']]], + ['vector_2eh_522',['vector.h',['../vector_8h.html',1,'']]], + ['vector_3c_20readonlysendertype_20_3e_523',['vector< ReadOnlySenderType >',['../structdlaf_1_1common_1_1internal_1_1vector.html',1,'dlaf::common::internal']]], + ['view_524',['View',['../structdlaf_1_1matrix_1_1internal_1_1_view.html',1,'dlaf::matrix::internal::View'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a3635af3164815f905a9fd33979a14a34',1,'dlaf::matrix::internal::View::View()']]], + ['views_2eh_525',['views.h',['../views_8h.html',1,'']]] ]; diff --git a/master/search/all_15.js b/master/search/all_15.js index ca9111353f..3701bc81d5 100644 --- a/master/search/all_15.js +++ b/master/search/all_15.js @@ -1,7 +1,7 @@ var searchData= [ - ['waitlocaltiles_522',['waitLocalTiles',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#abffc73bf079f7f7c23af3fca4dec329f',1,'dlaf::matrix::Matrix< const T, D >::waitLocalTiles()'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#ab731cfea1c5b07288bc3b34c59cb37d0',1,'dlaf::matrix::RetiledMatrix::waitLocalTiles()']]], - ['workspace_523',['WorkSpace',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space.html',1,'dlaf::eigensolver::internal']]], - ['workspacehost_524',['WorkSpaceHost',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space_host.html',1,'dlaf::eigensolver::internal']]], - ['workspacehostmirror_525',['WorkSpaceHostMirror',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space_host_mirror.html',1,'dlaf::eigensolver::internal']]] + ['waitlocaltiles_526',['waitLocalTiles',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#abffc73bf079f7f7c23af3fca4dec329f',1,'dlaf::matrix::Matrix< const T, D >']]], + ['workspace_527',['WorkSpace',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space.html',1,'dlaf::eigensolver::internal']]], + ['workspacehost_528',['WorkSpaceHost',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space_host.html',1,'dlaf::eigensolver::internal']]], + ['workspacehostmirror_529',['WorkSpaceHostMirror',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space_host_mirror.html',1,'dlaf::eigensolver::internal']]] ]; diff --git a/master/search/all_16.js b/master/search/all_16.js index 8d7a8e4270..75f0deea6c 100644 --- a/master/search/all_16.js +++ b/master/search/all_16.js @@ -1,9 +1,9 @@ var searchData= [ - ['_7ematrixmirror_526',['~MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#a2eeef61985e20a9218d55e76548acfb5',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::~MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#ae24915137508619315773c7a416f902f',1,'dlaf::matrix::MatrixMirror::~MatrixMirror()']]], - ['_7ememorychunk_527',['~MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html#a597aa1c075db44808f7e14c9ec03e50b',1,'dlaf::memory::MemoryChunk']]], - ['_7empi_5finit_528',['~mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html#a2413c9dc23983c5790cd0ebb3bb9e057',1,'dlaf::comm::mpi_init']]], - ['_7epanel_529',['~Panel',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#adab53d11d9904715b775df7c23ba6682',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['_7etile_530',['~Tile',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a66f6b8ef0f4d9168c12715d0fc1b16dc',1,'dlaf::matrix::Tile< const T, D >']]], - ['_7etype_5fhandler_531',['~type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a8fab81a168aa4f3441deb1a01f4a5171',1,'dlaf::comm::internal::type_handler']]] + ['_7ematrixmirror_530',['~MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#a2eeef61985e20a9218d55e76548acfb5',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::~MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#ae24915137508619315773c7a416f902f',1,'dlaf::matrix::MatrixMirror::~MatrixMirror()']]], + ['_7ememorychunk_531',['~MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html#a597aa1c075db44808f7e14c9ec03e50b',1,'dlaf::memory::MemoryChunk']]], + ['_7empi_5finit_532',['~mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html#a2413c9dc23983c5790cd0ebb3bb9e057',1,'dlaf::comm::mpi_init']]], + ['_7epanel_533',['~Panel',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#adab53d11d9904715b775df7c23ba6682',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['_7etile_534',['~Tile',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a66f6b8ef0f4d9168c12715d0fc1b16dc',1,'dlaf::matrix::Tile< const T, D >']]], + ['_7etype_5fhandler_535',['~type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a8fab81a168aa4f3441deb1a01f4a5171',1,'dlaf::comm::internal::type_handler']]] ]; diff --git a/master/search/all_3.js b/master/search/all_3.js index 1ac20e1cbb..1b3ce8f659 100644 --- a/master/search/all_3.js +++ b/master/search/all_3.js @@ -1,8 +1,8 @@ var searchData= [ - ['data_97',['data',['../classdlaf_1_1comm_1_1_message.html#a3684d7d2edac45401785fa9e8561eb73',1,'dlaf::comm::Message::data()'],['../structdlaf_1_1common_1_1_data_descriptor.html#a6f93fef2051c96e6b0100bcb50c280e2',1,'dlaf::common::DataDescriptor::data()']]], + ['data_97',['data',['../structdlaf_1_1common_1_1_data_descriptor.html#a6f93fef2051c96e6b0100bcb50c280e2',1,'dlaf::common::DataDescriptor::data()'],['../classdlaf_1_1comm_1_1_message.html#a3684d7d2edac45401785fa9e8561eb73',1,'dlaf::comm::Message::data()']]], ['data_2eh_98',['data.h',['../data_8h.html',1,'']]], - ['data_5f_99',['data_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a0aadac7bac5b0878d00305d21666c342',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::data_()'],['../classdlaf_1_1comm_1_1_message.html#a93caa3c42cfc16e9a27c65c37fa42c72',1,'dlaf::comm::Message::data_()']]], + ['data_5f_99',['data_',['../classdlaf_1_1comm_1_1_message.html#a93caa3c42cfc16e9a27c65c37fa42c72',1,'dlaf::comm::Message::data_()'],['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a0aadac7bac5b0878d00305d21666c342',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::data_()']]], ['data_5fblocksize_100',['data_blocksize',['../data_8h.html#aee766fd59c4a4d3c428082b37b54ea81',1,'dlaf::common']]], ['data_5fcount_101',['data_count',['../data_8h.html#a4b0d72a2fdf94917b5641c25c86bcf18',1,'dlaf::common']]], ['data_5fdescriptor_2eh_102',['data_descriptor.h',['../data__descriptor_8h.html',1,'']]], @@ -14,7 +14,7 @@ var searchData= ['data_5ftraits_3c_20buffer_3c_20t_20_3e_20_3e_108',['data_traits< Buffer< T > >',['../structdlaf_1_1common_1_1data__traits_3_01_buffer_3_01_t_01_4_01_4.html',1,'dlaf::common']]], ['data_5ftraits_3c_20datadescriptor_3c_20t_20_3e_20_3e_109',['data_traits< DataDescriptor< T > >',['../structdlaf_1_1common_1_1data__traits_3_01_data_descriptor_3_01_t_01_4_01_4.html',1,'dlaf::common']]], ['data_5ftraits_3c_20datadescriptor_3c_20t_5bn_5d_3e_20_3e_110',['data_traits< DataDescriptor< T[N]> >',['../structdlaf_1_1common_1_1data__traits_3_01_data_descriptor_3_01_t_0f_n_0e_4_01_4.html',1,'dlaf::common']]], - ['datadescriptor_111',['DataDescriptor',['../structdlaf_1_1common_1_1_data_descriptor_3_01_t_0f_n_0e_4.html#acef674017777de4a735258a476e95e47',1,'dlaf::common::DataDescriptor< T[N]>::DataDescriptor()'],['../structdlaf_1_1common_1_1_data_descriptor.html#acb58ba4559ea0a6022a45bb781dccc6c',1,'dlaf::common::DataDescriptor::DataDescriptor(const Data &data)'],['../structdlaf_1_1common_1_1_data_descriptor.html#acd73354fd77df45c9b70ab034fc1cd61',1,'dlaf::common::DataDescriptor::DataDescriptor(T *ptr, SizeType num_blocks, SizeType blocksize, SizeType stride) noexcept'],['../structdlaf_1_1common_1_1_data_descriptor.html#a8c969a23a32e1cf2ae19a7811ae394e4',1,'dlaf::common::DataDescriptor::DataDescriptor(T *ptr, SizeType n) noexcept'],['../structdlaf_1_1common_1_1_data_descriptor.html#a87ca3689fd590f74e636317987af17bd',1,'dlaf::common::DataDescriptor::DataDescriptor()'],['../structdlaf_1_1common_1_1_data_descriptor.html',1,'dlaf::common::DataDescriptor< T >']]], + ['datadescriptor_111',['DataDescriptor',['../structdlaf_1_1common_1_1_data_descriptor.html#a8c969a23a32e1cf2ae19a7811ae394e4',1,'dlaf::common::DataDescriptor::DataDescriptor(T *ptr, SizeType n) noexcept'],['../structdlaf_1_1common_1_1_data_descriptor.html#acd73354fd77df45c9b70ab034fc1cd61',1,'dlaf::common::DataDescriptor::DataDescriptor(T *ptr, SizeType num_blocks, SizeType blocksize, SizeType stride) noexcept'],['../structdlaf_1_1common_1_1_data_descriptor.html#acb58ba4559ea0a6022a45bb781dccc6c',1,'dlaf::common::DataDescriptor::DataDescriptor(const Data &data)'],['../structdlaf_1_1common_1_1_data_descriptor_3_01_t_0f_n_0e_4.html#acef674017777de4a735258a476e95e47',1,'dlaf::common::DataDescriptor< T[N]>::DataDescriptor()'],['../structdlaf_1_1common_1_1_data_descriptor.html#a87ca3689fd590f74e636317987af17bd',1,'dlaf::common::DataDescriptor::DataDescriptor()'],['../structdlaf_1_1common_1_1_data_descriptor.html',1,'dlaf::common::DataDescriptor< T >']]], ['datadescriptor_3c_20t_5bn_5d_3e_112',['DataDescriptor< T[N]>',['../structdlaf_1_1common_1_1_data_descriptor_3_01_t_0f_n_0e_4.html',1,'dlaf::common']]], ['datatypes_2eh_113',['datatypes.h',['../datatypes_8h.html',1,'']]], ['defaultbackend_114',['DefaultBackend',['../structdlaf_1_1_default_backend.html',1,'dlaf']]], @@ -27,12 +27,13 @@ var searchData= ['dist_5fmatrix_5f_121',['dist_matrix_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ab991456b59d36ac84dd9277edefc42ac',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], ['distancetoadjacenttile_122',['distanceToAdjacentTile',['../classdlaf_1_1matrix_1_1_distribution.html#aa0b67d70b8b3e6f74c53d1062546b8ed',1,'dlaf::matrix::Distribution']]], ['distindexing_123',['DistIndexing',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_dist_indexing.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['distribution_124',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html#a97b363d57f32add582ce1774a6b9a147',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a47b322556c713fd38223538f50f58e14',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalTileIndex &tile_offset, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a51c38487398212457417f784dd3d2594',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const TileElementSize &tile_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a012f931d7acdf168a42445bcbbce5dd3',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const TileElementSize &tile_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalTileIndex &tile_offset, const GlobalElementIndex &element_offset={0, 0})']]], + ['distribution_124',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html#a97b363d57f32add582ce1774a6b9a147',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a012f931d7acdf168a42445bcbbce5dd3',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const TileElementSize &tile_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalTileIndex &tile_offset, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a51c38487398212457417f784dd3d2594',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const TileElementSize &tile_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a47b322556c713fd38223538f50f58e14',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalTileIndex &tile_offset, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a146465cb553b8a9d9f32bf63ff7e7375',1,'dlaf::matrix::Distribution::Distribution(const LocalElementSize &size, const TileElementSize &block_size, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#aacf1f8d3d02d9c70bf0bffe0725b5d92',1,'dlaf::matrix::Distribution::Distribution() noexcept']]], ['distribution_125',['distribution',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a28967ff509064a6e04cfc8b2d35af402',1,'dlaf::matrix::internal::MatrixBase']]], - ['distribution_126',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html#a146465cb553b8a9d9f32bf63ff7e7375',1,'dlaf::matrix::Distribution::Distribution(const LocalElementSize &size, const TileElementSize &block_size, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#aacf1f8d3d02d9c70bf0bffe0725b5d92',1,'dlaf::matrix::Distribution::Distribution() noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html',1,'dlaf::matrix::Distribution']]], + ['distribution_126',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html',1,'dlaf::matrix']]], ['distribution_2eh_127',['distribution.h',['../distribution_8h.html',1,'']]], ['distworkspacehostmirror_128',['DistWorkSpaceHostMirror',['../structdlaf_1_1eigensolver_1_1internal_1_1_dist_work_space_host_mirror.html',1,'dlaf::eigensolver::internal']]], ['dlaf_5fmake_5fcallable_5fobject_129',['DLAF_MAKE_CALLABLE_OBJECT',['../callable__object_8h.html#a5aafa5a9c65ad07da48410427f4825b2',1,'callable_object.h']]], ['dlaf_5fmake_5fsender_5falgorithm_5foverloads_130',['DLAF_MAKE_SENDER_ALGORITHM_OVERLOADS',['../copy__tile_8h.html#a57758eefb97e5d9fef7405d84c90d867',1,'dlaf::matrix']]], - ['duplicatenocopy_131',['DuplicateNoCopy',['../structdlaf_1_1matrix_1_1_duplicate_no_copy.html',1,'dlaf::matrix']]] + ['done_131',['done',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#aaac68e6ec599e12aac6e7e4bd32570a8',1,'dlaf::matrix::Matrix< const T, D >::done(const LocalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#ade093c346a4df7e85c94b71fd86eaec2',1,'dlaf::matrix::Matrix< const T, D >::done(const GlobalTileIndex &index) noexcept']]], + ['duplicatenocopy_132',['DuplicateNoCopy',['../structdlaf_1_1matrix_1_1_duplicate_no_copy.html',1,'dlaf::matrix']]] ]; diff --git a/master/search/all_4.js b/master/search/all_4.js index c3913f62c4..a314b6685e 100644 --- a/master/search/all_4.js +++ b/master/search/all_4.js @@ -1,19 +1,19 @@ var searchData= [ - ['eigensolver_132',['Eigensolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_eigensolver.html',1,'dlaf::eigensolver::internal']]], - ['eigensolver_133',['eigensolver',['../eigensolver_2eigensolver_8h.html#ad43e9ff356aafef70396fdc6b6d3a1f2',1,'dlaf::eigensolver::eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat)'],['../eigensolver_2eigensolver_8h.html#a2b99692c7224f63a547cc6686d732490',1,'dlaf::eigensolver::eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../eigensolver_2eigensolver_8h.html#a94fb08e9087b48ae5fe19b59c9e28db8',1,'dlaf::eigensolver::eigensolver(blas::Uplo uplo, Matrix< T, D > &mat)'],['../eigensolver_2eigensolver_8h.html#a7aeed252a1783c24f778eb250d0efd72',1,'dlaf::eigensolver::eigensolver(blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)']]], - ['eigensolver_2eh_134',['eigensolver.h',['../eigensolver_2eigensolver_8h.html',1,'(Global Namespace)'],['../eigensolver_8h.html',1,'(Global Namespace)']]], - ['eigensolverresult_135',['EigensolverResult',['../structdlaf_1_1eigensolver_1_1_eigensolver_result.html',1,'dlaf::eigensolver']]], - ['elementfromtileandtileelement_136',['elementFromTileAndTileElement',['../util__distribution_8h.html#a9eab8f0da313765c25ddd2eec12665e0',1,'dlaf::util::matrix']]], - ['emptyenv_137',['EmptyEnv',['../structdlaf_1_1internal_1_1_empty_env.html',1,'dlaf::internal']]], - ['enable_5fif_5fconvertible_5ft_138',['enable_if_convertible_t',['../traits_8h.html#a234bd5c290fe869e9856e7e2d3667baf',1,'dlaf']]], - ['enable_5fif_5fsignature_5ft_139',['enable_if_signature_t',['../traits_8h.html#ada84be4e014ed8f3d91175b3856104e5',1,'dlaf']]], - ['end_5f_140',['end_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a6d52e171776a86d6a42b2e0cc7a26e7b',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['enum_5foutput_2eh_141',['enum_output.h',['../blas_2enum__output_8h.html',1,'']]], - ['equal_5fblocksize_142',['equal_blocksize',['../util__matrix_8h.html#a0bb54ace6dfcb245ec387c1af75e63fe',1,'dlaf::matrix']]], - ['equal_5fdistributions_143',['equal_distributions',['../util__matrix_8h.html#a8af8a409c1af4f1c7cdbed9b4f6b24a2',1,'dlaf::matrix']]], - ['equal_5fprocess_5fgrid_144',['equal_process_grid',['../util__matrix_8h.html#a7a41b22af7fe104052668cc02d5a624c',1,'dlaf::matrix']]], - ['equal_5fsize_145',['equal_size',['../util__matrix_8h.html#ada06af4450b10ed70a5830c0d8db0302',1,'dlaf::matrix']]], - ['error_2eh_146',['error.h',['../communication_2error_8h.html',1,'']]], - ['external_5f_147',['external_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a785fd3e6c916a2ae30efb17f3449c230',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] + ['eigensolver_133',['Eigensolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_eigensolver.html',1,'dlaf::eigensolver::internal']]], + ['eigensolver_134',['eigensolver',['../eigensolver_2eigensolver_8h.html#ad43e9ff356aafef70396fdc6b6d3a1f2',1,'dlaf::eigensolver::eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat)'],['../eigensolver_2eigensolver_8h.html#a2b99692c7224f63a547cc6686d732490',1,'dlaf::eigensolver::eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../eigensolver_2eigensolver_8h.html#a94fb08e9087b48ae5fe19b59c9e28db8',1,'dlaf::eigensolver::eigensolver(blas::Uplo uplo, Matrix< T, D > &mat)'],['../eigensolver_2eigensolver_8h.html#a7aeed252a1783c24f778eb250d0efd72',1,'dlaf::eigensolver::eigensolver(blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)']]], + ['eigensolver_2eh_135',['eigensolver.h',['../eigensolver_2eigensolver_8h.html',1,'(Global Namespace)'],['../eigensolver_8h.html',1,'(Global Namespace)']]], + ['eigensolverresult_136',['EigensolverResult',['../structdlaf_1_1eigensolver_1_1_eigensolver_result.html',1,'dlaf::eigensolver']]], + ['elementfromtileandtileelement_137',['elementFromTileAndTileElement',['../util__distribution_8h.html#a9eab8f0da313765c25ddd2eec12665e0',1,'dlaf::util::matrix']]], + ['emptyenv_138',['EmptyEnv',['../structdlaf_1_1internal_1_1_empty_env.html',1,'dlaf::internal']]], + ['enable_5fif_5fconvertible_5ft_139',['enable_if_convertible_t',['../traits_8h.html#a234bd5c290fe869e9856e7e2d3667baf',1,'dlaf']]], + ['enable_5fif_5fsignature_5ft_140',['enable_if_signature_t',['../traits_8h.html#ada84be4e014ed8f3d91175b3856104e5',1,'dlaf']]], + ['end_5f_141',['end_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a6d52e171776a86d6a42b2e0cc7a26e7b',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['enum_5foutput_2eh_142',['enum_output.h',['../blas_2enum__output_8h.html',1,'']]], + ['equal_5fblocksize_143',['equal_blocksize',['../util__matrix_8h.html#a0bb54ace6dfcb245ec387c1af75e63fe',1,'dlaf::matrix']]], + ['equal_5fdistributions_144',['equal_distributions',['../util__matrix_8h.html#a8af8a409c1af4f1c7cdbed9b4f6b24a2',1,'dlaf::matrix']]], + ['equal_5fprocess_5fgrid_145',['equal_process_grid',['../util__matrix_8h.html#a7a41b22af7fe104052668cc02d5a624c',1,'dlaf::matrix']]], + ['equal_5fsize_146',['equal_size',['../util__matrix_8h.html#ada06af4450b10ed70a5830c0d8db0302',1,'dlaf::matrix']]], + ['error_2eh_147',['error.h',['../communication_2error_8h.html',1,'']]], + ['external_5f_148',['external_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a785fd3e6c916a2ae30efb17f3449c230',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] ]; diff --git a/master/search/all_5.js b/master/search/all_5.js index 9b674bf340..fdaf360602 100644 --- a/master/search/all_5.js +++ b/master/search/all_5.js @@ -1,10 +1,10 @@ var searchData= [ - ['factorization_2eh_148',['factorization.h',['../factorization_8h.html',1,'']]], - ['finalize_149',['finalize',['../init_8h.html#abf5448f6d8cff5d741a7680123a7082c',1,'dlaf']]], - ['format_5fshort_2eh_150',['format_short.h',['../format__short_8h.html',1,'']]], - ['formatshort_151',['FormatShort',['../structdlaf_1_1internal_1_1_format_short.html',1,'dlaf::internal']]], - ['fullcommunicator_152',['fullCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#a1cb5842312d3e88e298819ea192d286d',1,'dlaf::comm::CommunicatorGrid']]], - ['fullindex_153',['fullIndex',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a709e5ba882750052d23a5a8eb1eff21a',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['functions_5fsync_2eh_154',['functions_sync.h',['../functions__sync_8h.html',1,'']]] + ['factorization_2eh_149',['factorization.h',['../factorization_8h.html',1,'']]], + ['finalize_150',['finalize',['../init_8h.html#abf5448f6d8cff5d741a7680123a7082c',1,'dlaf']]], + ['format_5fshort_2eh_151',['format_short.h',['../format__short_8h.html',1,'']]], + ['formatshort_152',['FormatShort',['../structdlaf_1_1internal_1_1_format_short.html',1,'dlaf::internal']]], + ['fullcommunicator_153',['fullCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#a1cb5842312d3e88e298819ea192d286d',1,'dlaf::comm::CommunicatorGrid']]], + ['fullindex_154',['fullIndex',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a709e5ba882750052d23a5a8eb1eff21a',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['functions_5fsync_2eh_155',['functions_sync.h',['../functions__sync_8h.html',1,'']]] ]; diff --git a/master/search/all_6.js b/master/search/all_6.js index 249715a933..20bade51eb 100644 --- a/master/search/all_6.js +++ b/master/search/all_6.js @@ -1,31 +1,31 @@ var searchData= [ - ['gemm_155',['gemm',['../blas_2tile_8h.html#a8e495a0a86ddf2f997ca8a7c33d106a2',1,'dlaf::tile::gemm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#afed112b0e26de470da2469d43866918d',1,'dlaf::tile::gemm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a1de961c6e1c01ee6a5f0039ecf51b162',1,'dlaf::tile::gemm(const blas::Op op_a, const blas::Op op_b, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const T beta, const Tile< T, D > &c)']]], - ['gemmsizes_156',['gemmSizes',['../structdlaf_1_1tile_1_1internal_1_1gemm_sizes.html',1,'dlaf::tile::internal']]], - ['gen_5feigensolver_2eh_157',['gen_eigensolver.h',['../gen__eigensolver_8h.html',1,'']]], - ['gen_5fto_5fstd_2eh_158',['gen_to_std.h',['../gen__to__std_8h.html',1,'']]], - ['geneigensolver_159',['GenEigensolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_gen_eigensolver.html',1,'dlaf::eigensolver::internal']]], - ['geneigensolver_160',['genEigensolver',['../gen__eigensolver_8h.html#ad60b7fb5135b8ea685d781e4e4165c47',1,'dlaf::eigensolver::genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b)'],['../gen__eigensolver_8h.html#ae119b500260a60dcabdd237e461a12dd',1,'dlaf::eigensolver::genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../gen__eigensolver_8h.html#a9d80f2453da0de6a6ad73c55165f7f52',1,'dlaf::eigensolver::genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../gen__eigensolver_8h.html#a7578fbe49ba90d7ef916ac81860c3858',1,'dlaf::eigensolver::genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b)']]], - ['general_2eh_161',['general.h',['../multiplication_2general_8h.html',1,'(Global Namespace)'],['../permutations_2general_8h.html',1,'(Global Namespace)']]], - ['generalsub_162',['GeneralSub',['../structdlaf_1_1multiplication_1_1internal_1_1_general_sub.html',1,'dlaf::multiplication::internal']]], - ['generalsubmatrix_163',['generalSubMatrix',['../multiplication_2general_8h.html#a45f04fd24259c413a07a210a17b91b93',1,'dlaf::multiplication::generalSubMatrix(const SizeType a, const SizeType b, const blas::Op opA, const blas::Op opB, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)'],['../multiplication_2general_8h.html#a0aa788470aca6ca60202513251578a36',1,'dlaf::multiplication::generalSubMatrix([[maybe_unused]] comm::CommunicatorGrid grid, common::Pipeline< comm::Communicator > &row_task_chain, common::Pipeline< comm::Communicator > &col_task_chain, const SizeType a, const SizeType b, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)']]], - ['generic_5finteger_5fop_164',['generic_integer_op',['../util__math_8h.html#a4a1c1045d16c92d7fef9c553ae8e6c90',1,'dlaf::util::internal']]], - ['gentostd_165',['GenToStd',['../structdlaf_1_1eigensolver_1_1internal_1_1_gen_to_std.html',1,'dlaf::eigensolver::internal']]], - ['gentostd_166',['genToStd',['../gen__to__std_8h.html#a1c178c1c870afbc1c4928a4dbe078aa4',1,'dlaf::eigensolver::genToStd(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)'],['../gen__to__std_8h.html#a8f0b0f0094c0185dc428f69f8506ba8f',1,'dlaf::eigensolver::genToStd(blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)']]], - ['get_167',['get',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#a8b053672d5ecf63c2af3a3c23677278b',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#ad2cba1bf3133441bc31d960d2681cf11',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#aa0d470acb7c2f97a7192c06fceef7b32',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#ad5eaeb945174c2c237a672dd4191582a',1,'dlaf::matrix::MatrixMirror::get()'],['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a23701a66458fad55204f90b84674a8e2',1,'dlaf::common::internal::basic_coords::get()']]], - ['getheight_168',['getHeight',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ad59bcb70b44cf2a4a61119d8ce4eb5db',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['getoptionsdescription_169',['getOptionsDescription',['../init_8h.html#a6f4711b407727e4b9317d252b451bd89',1,'dlaf']]], - ['getsource_170',['getSource',['../classdlaf_1_1matrix_1_1_matrix_mirror.html#a2fb277eb0664e0e6ed0e8a011416c7c7',1,'dlaf::matrix::MatrixMirror::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#a424e6b85bb0167015efbb7522a31d7fa',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a348f9f6d9af5df67ef6179e4f600850e',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#af6ab705620e03736d36433f0846470fc',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::getSource()']]], - ['getter_5frandom_171',['getter_random',['../classdlaf_1_1matrix_1_1util_1_1internal_1_1getter__random.html',1,'dlaf::matrix::util::internal']]], - ['getter_5frandom_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_172',['getter_random< std::complex< T > >',['../classdlaf_1_1matrix_1_1util_1_1internal_1_1getter__random_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::matrix::util::internal']]], - ['getwidth_173',['getWidth',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa34a2e93ba64d958df32428f3928ccba',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['givensrotation_174',['GivensRotation',['../structdlaf_1_1eigensolver_1_1internal_1_1_givens_rotation.html',1,'dlaf::eigensolver::internal']]], - ['globalelementfromglobaltileandtileelement_175',['globalElementFromGlobalTileAndTileElement',['../classdlaf_1_1matrix_1_1_distribution.html#aed7a91383a0146b02008c6d6bfed2850',1,'dlaf::matrix::Distribution']]], - ['globalelementfromlocaltileandtileelement_176',['globalElementFromLocalTileAndTileElement',['../classdlaf_1_1matrix_1_1_distribution.html#a36c1a5962e433c773a924d8b5bdda594',1,'dlaf::matrix::Distribution']]], - ['globalelementindex_177',['globalElementIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a20234ff229ec8872625293b28c7d47f5',1,'dlaf::matrix::Distribution']]], - ['globaltileelementdistance_178',['globalTileElementDistance',['../classdlaf_1_1matrix_1_1_distribution.html#a9f57d9074212c024d39f0e971bc9e5db',1,'dlaf::matrix::Distribution']]], - ['globaltilefromglobalelement_179',['globalTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#ae63d662d123b38f1150b553b2cdb259d',1,'dlaf::matrix::Distribution']]], - ['globaltilefromlocaltile_180',['globalTileFromLocalTile',['../classdlaf_1_1matrix_1_1_distribution.html#add4b487f6603fb6099dce376ee2e8c3c',1,'dlaf::matrix::Distribution::globalTileFromLocalTile()'],['../util__distribution_8h.html#a3fa80781a6b118e00015b9751e2f6ead',1,'dlaf::util::matrix::globalTileFromLocalTile()']]], - ['globaltileindex_181',['globalTileIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a7a0027799c7c187d8f7135a1242ca3f5',1,'dlaf::matrix::Distribution::globalTileIndex(const LocalTileIndex &local_tile) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ae49ef916d77bf8de7e8e4710a1034299',1,'dlaf::matrix::Distribution::globalTileIndex(const GlobalElementIndex &global_element) const noexcept']]], - ['globaltilelinearindex_182',['globalTileLinearIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a0b174a696d5fe710ecb2bc081081c137',1,'dlaf::matrix::Distribution']]] + ['gemm_156',['gemm',['../blas_2tile_8h.html#a8e495a0a86ddf2f997ca8a7c33d106a2',1,'dlaf::tile::gemm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#afed112b0e26de470da2469d43866918d',1,'dlaf::tile::gemm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a1de961c6e1c01ee6a5f0039ecf51b162',1,'dlaf::tile::gemm(const blas::Op op_a, const blas::Op op_b, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const T beta, const Tile< T, D > &c)']]], + ['gemmsizes_157',['gemmSizes',['../structdlaf_1_1tile_1_1internal_1_1gemm_sizes.html',1,'dlaf::tile::internal']]], + ['gen_5feigensolver_2eh_158',['gen_eigensolver.h',['../gen__eigensolver_8h.html',1,'']]], + ['gen_5fto_5fstd_2eh_159',['gen_to_std.h',['../gen__to__std_8h.html',1,'']]], + ['geneigensolver_160',['GenEigensolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_gen_eigensolver.html',1,'dlaf::eigensolver::internal']]], + ['geneigensolver_161',['genEigensolver',['../gen__eigensolver_8h.html#ad60b7fb5135b8ea685d781e4e4165c47',1,'dlaf::eigensolver::genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b)'],['../gen__eigensolver_8h.html#ae119b500260a60dcabdd237e461a12dd',1,'dlaf::eigensolver::genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../gen__eigensolver_8h.html#a9d80f2453da0de6a6ad73c55165f7f52',1,'dlaf::eigensolver::genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../gen__eigensolver_8h.html#a7578fbe49ba90d7ef916ac81860c3858',1,'dlaf::eigensolver::genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b)']]], + ['general_2eh_162',['general.h',['../multiplication_2general_8h.html',1,'(Global Namespace)'],['../permutations_2general_8h.html',1,'(Global Namespace)']]], + ['generalsub_163',['GeneralSub',['../structdlaf_1_1multiplication_1_1internal_1_1_general_sub.html',1,'dlaf::multiplication::internal']]], + ['generalsubmatrix_164',['generalSubMatrix',['../multiplication_2general_8h.html#a45f04fd24259c413a07a210a17b91b93',1,'dlaf::multiplication::generalSubMatrix(const SizeType a, const SizeType b, const blas::Op opA, const blas::Op opB, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)'],['../multiplication_2general_8h.html#a0aa788470aca6ca60202513251578a36',1,'dlaf::multiplication::generalSubMatrix([[maybe_unused]] comm::CommunicatorGrid grid, common::Pipeline< comm::Communicator > &row_task_chain, common::Pipeline< comm::Communicator > &col_task_chain, const SizeType a, const SizeType b, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)']]], + ['generic_5finteger_5fop_165',['generic_integer_op',['../util__math_8h.html#a4a1c1045d16c92d7fef9c553ae8e6c90',1,'dlaf::util::internal']]], + ['gentostd_166',['GenToStd',['../structdlaf_1_1eigensolver_1_1internal_1_1_gen_to_std.html',1,'dlaf::eigensolver::internal']]], + ['gentostd_167',['genToStd',['../gen__to__std_8h.html#a1c178c1c870afbc1c4928a4dbe078aa4',1,'dlaf::eigensolver::genToStd(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)'],['../gen__to__std_8h.html#a8f0b0f0094c0185dc428f69f8506ba8f',1,'dlaf::eigensolver::genToStd(blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)']]], + ['get_168',['get',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#a8b053672d5ecf63c2af3a3c23677278b',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#ad2cba1bf3133441bc31d960d2681cf11',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#aa0d470acb7c2f97a7192c06fceef7b32',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#ad5eaeb945174c2c237a672dd4191582a',1,'dlaf::matrix::MatrixMirror::get()'],['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a23701a66458fad55204f90b84674a8e2',1,'dlaf::common::internal::basic_coords::get()']]], + ['getheight_169',['getHeight',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ad59bcb70b44cf2a4a61119d8ce4eb5db',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['getoptionsdescription_170',['getOptionsDescription',['../init_8h.html#a6f4711b407727e4b9317d252b451bd89',1,'dlaf']]], + ['getsource_171',['getSource',['../classdlaf_1_1matrix_1_1_matrix_mirror.html#a2fb277eb0664e0e6ed0e8a011416c7c7',1,'dlaf::matrix::MatrixMirror::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#a424e6b85bb0167015efbb7522a31d7fa',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a348f9f6d9af5df67ef6179e4f600850e',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#af6ab705620e03736d36433f0846470fc',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::getSource()']]], + ['getter_5frandom_172',['getter_random',['../classdlaf_1_1matrix_1_1util_1_1internal_1_1getter__random.html',1,'dlaf::matrix::util::internal']]], + ['getter_5frandom_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_173',['getter_random< std::complex< T > >',['../classdlaf_1_1matrix_1_1util_1_1internal_1_1getter__random_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::matrix::util::internal']]], + ['getwidth_174',['getWidth',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa34a2e93ba64d958df32428f3928ccba',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['givensrotation_175',['GivensRotation',['../structdlaf_1_1eigensolver_1_1internal_1_1_givens_rotation.html',1,'dlaf::eigensolver::internal']]], + ['globalelementfromglobaltileandtileelement_176',['globalElementFromGlobalTileAndTileElement',['../classdlaf_1_1matrix_1_1_distribution.html#aed7a91383a0146b02008c6d6bfed2850',1,'dlaf::matrix::Distribution']]], + ['globalelementfromlocaltileandtileelement_177',['globalElementFromLocalTileAndTileElement',['../classdlaf_1_1matrix_1_1_distribution.html#a36c1a5962e433c773a924d8b5bdda594',1,'dlaf::matrix::Distribution']]], + ['globalelementindex_178',['globalElementIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a20234ff229ec8872625293b28c7d47f5',1,'dlaf::matrix::Distribution']]], + ['globaltileelementdistance_179',['globalTileElementDistance',['../classdlaf_1_1matrix_1_1_distribution.html#a9f57d9074212c024d39f0e971bc9e5db',1,'dlaf::matrix::Distribution']]], + ['globaltilefromglobalelement_180',['globalTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#ae63d662d123b38f1150b553b2cdb259d',1,'dlaf::matrix::Distribution']]], + ['globaltilefromlocaltile_181',['globalTileFromLocalTile',['../classdlaf_1_1matrix_1_1_distribution.html#add4b487f6603fb6099dce376ee2e8c3c',1,'dlaf::matrix::Distribution::globalTileFromLocalTile()'],['../util__distribution_8h.html#a3fa80781a6b118e00015b9751e2f6ead',1,'dlaf::util::matrix::globalTileFromLocalTile()']]], + ['globaltileindex_182',['globalTileIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a7a0027799c7c187d8f7135a1242ca3f5',1,'dlaf::matrix::Distribution::globalTileIndex(const LocalTileIndex &local_tile) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ae49ef916d77bf8de7e8e4710a1034299',1,'dlaf::matrix::Distribution::globalTileIndex(const GlobalElementIndex &global_element) const noexcept']]], + ['globaltilelinearindex_183',['globalTileLinearIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a0b174a696d5fe710ecb2bc081081c137',1,'dlaf::matrix::Distribution']]] ]; diff --git a/master/search/all_7.js b/master/search/all_7.js index e2e4ab3a7c..a2bc68adcb 100644 --- a/master/search/all_7.js +++ b/master/search/all_7.js @@ -1,21 +1,21 @@ var searchData= [ - ['hegst_183',['hegst',['../lapack_2tile_8h.html#aaf274e91e36453e7994dd1f2a2813681',1,'dlaf::tile']]], - ['helpers_184',['Helpers',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers.html',1,'dlaf::eigensolver::internal::bt_red_band::Helpers< B >'],['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers.html',1,'dlaf::factorization::internal::tfactor_l::Helpers< backend, device, T >']]], - ['helpers_3c_20backend_3a_3agpu_20_3e_185',['Helpers< Backend::GPU >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers_3_01_backend_1_1_g_p_u_01_4.html',1,'dlaf::eigensolver::internal::bt_red_band']]], - ['helpers_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_186',['Helpers< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers_3_01_backend_1_1_g_p_u_00_01_device_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::factorization::internal::tfactor_l']]], - ['helpers_3c_20backend_3a_3amc_20_3e_187',['Helpers< Backend::MC >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers_3_01_backend_1_1_m_c_01_4.html',1,'dlaf::eigensolver::internal::bt_red_band']]], - ['helpers_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_188',['Helpers< Backend::MC, Device::CPU, T >',['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::factorization::internal::tfactor_l']]], - ['hemm_189',['hemm',['../blas_2tile_8h.html#a8887bb899e49f4ea3acce81474f5965d',1,'dlaf::tile::hemm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#ab7031513f6c7338a1f3cea71784d0e9a',1,'dlaf::tile::hemm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#aeec422be0ef5abea8ab6f1a8ae40c8ef',1,'dlaf::tile::hemm(const blas::Side side, const blas::Uplo uplo, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const T beta, const Tile< T, D > &c)']]], - ['hemmsizes_190',['hemmSizes',['../structdlaf_1_1tile_1_1internal_1_1hemm_sizes.html',1,'dlaf::tile::internal']]], - ['her2k_191',['her2k',['../blas_2tile_8h.html#a45e96a215d6f3a7f64ccf05d108087e7',1,'dlaf::tile::her2k(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#aafcb69907e7154cb12e0e3121534310a',1,'dlaf::tile::her2k(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a36642d0aeb16e71f10fd289985f4a03f',1,'dlaf::tile::her2k(const blas::Uplo uplo, const blas::Op op, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const BaseType< T > beta, const Tile< T, D > &c)']]], - ['her2ksizes_192',['her2kSizes',['../structdlaf_1_1tile_1_1internal_1_1her2k_sizes.html',1,'dlaf::tile::internal']]], - ['herk_193',['herk',['../blas_2tile_8h.html#a72ab20a0ba5ba199c1ec18f8a31ee2eb',1,'dlaf::tile::herk(const blas::Uplo uplo, const blas::Op op, const BaseType< T > alpha, const Tile< const T, D > &a, const BaseType< T > beta, const Tile< T, D > &c)'],['../blas_2tile_8h.html#a8a4f39f0b26f02bc094e531add909e6e',1,'dlaf::tile::herk(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#aa8fd52359e36023fb6418834d4b27fcd',1,'dlaf::tile::herk(const dlaf::internal::Policy< B > &p)']]], - ['herksizes_194',['herkSizes',['../structdlaf_1_1tile_1_1internal_1_1herk_sizes.html',1,'dlaf::tile::internal']]], - ['hermitian_195',['Hermitian',['../structdlaf_1_1multiplication_1_1internal_1_1_hermitian.html',1,'dlaf::multiplication::internal']]], - ['hermitian_196',['hermitian',['../hermitian_8h.html#a117a0b955e26513fecdc562da0f96922',1,'dlaf::multiplication::hermitian(blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)'],['../hermitian_8h.html#a3a63ff51bec2e9383f3dfd8b2914673e',1,'dlaf::multiplication::hermitian(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)']]], - ['hermitian_2eh_197',['hermitian.h',['../hermitian_8h.html',1,'']]], - ['hhmanager_198',['HHManager',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['hhmanager_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_199',['HHManager< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager_3_01_backend_1_1_g_p_u_00_01_device_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['hhmanager_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_200',['HHManager< Backend::MC, Device::CPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]] + ['hegst_184',['hegst',['../lapack_2tile_8h.html#aaf274e91e36453e7994dd1f2a2813681',1,'dlaf::tile']]], + ['helpers_185',['Helpers',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers.html',1,'dlaf::eigensolver::internal::bt_red_band::Helpers< B >'],['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers.html',1,'dlaf::factorization::internal::tfactor_l::Helpers< backend, device, T >']]], + ['helpers_3c_20backend_3a_3agpu_20_3e_186',['Helpers< Backend::GPU >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers_3_01_backend_1_1_g_p_u_01_4.html',1,'dlaf::eigensolver::internal::bt_red_band']]], + ['helpers_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_187',['Helpers< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers_3_01_backend_1_1_g_p_u_00_01_device_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::factorization::internal::tfactor_l']]], + ['helpers_3c_20backend_3a_3amc_20_3e_188',['Helpers< Backend::MC >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers_3_01_backend_1_1_m_c_01_4.html',1,'dlaf::eigensolver::internal::bt_red_band']]], + ['helpers_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_189',['Helpers< Backend::MC, Device::CPU, T >',['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::factorization::internal::tfactor_l']]], + ['hemm_190',['hemm',['../blas_2tile_8h.html#a8887bb899e49f4ea3acce81474f5965d',1,'dlaf::tile::hemm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#ab7031513f6c7338a1f3cea71784d0e9a',1,'dlaf::tile::hemm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#aeec422be0ef5abea8ab6f1a8ae40c8ef',1,'dlaf::tile::hemm(const blas::Side side, const blas::Uplo uplo, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const T beta, const Tile< T, D > &c)']]], + ['hemmsizes_191',['hemmSizes',['../structdlaf_1_1tile_1_1internal_1_1hemm_sizes.html',1,'dlaf::tile::internal']]], + ['her2k_192',['her2k',['../blas_2tile_8h.html#a45e96a215d6f3a7f64ccf05d108087e7',1,'dlaf::tile::her2k(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#aafcb69907e7154cb12e0e3121534310a',1,'dlaf::tile::her2k(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a36642d0aeb16e71f10fd289985f4a03f',1,'dlaf::tile::her2k(const blas::Uplo uplo, const blas::Op op, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const BaseType< T > beta, const Tile< T, D > &c)']]], + ['her2ksizes_193',['her2kSizes',['../structdlaf_1_1tile_1_1internal_1_1her2k_sizes.html',1,'dlaf::tile::internal']]], + ['herk_194',['herk',['../blas_2tile_8h.html#a72ab20a0ba5ba199c1ec18f8a31ee2eb',1,'dlaf::tile::herk(const blas::Uplo uplo, const blas::Op op, const BaseType< T > alpha, const Tile< const T, D > &a, const BaseType< T > beta, const Tile< T, D > &c)'],['../blas_2tile_8h.html#a8a4f39f0b26f02bc094e531add909e6e',1,'dlaf::tile::herk(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#aa8fd52359e36023fb6418834d4b27fcd',1,'dlaf::tile::herk(const dlaf::internal::Policy< B > &p)']]], + ['herksizes_195',['herkSizes',['../structdlaf_1_1tile_1_1internal_1_1herk_sizes.html',1,'dlaf::tile::internal']]], + ['hermitian_196',['Hermitian',['../structdlaf_1_1multiplication_1_1internal_1_1_hermitian.html',1,'dlaf::multiplication::internal']]], + ['hermitian_197',['hermitian',['../hermitian_8h.html#a117a0b955e26513fecdc562da0f96922',1,'dlaf::multiplication::hermitian(blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)'],['../hermitian_8h.html#a3a63ff51bec2e9383f3dfd8b2914673e',1,'dlaf::multiplication::hermitian(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)']]], + ['hermitian_2eh_198',['hermitian.h',['../hermitian_8h.html',1,'']]], + ['hhmanager_199',['HHManager',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['hhmanager_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_200',['HHManager< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager_3_01_backend_1_1_g_p_u_00_01_device_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['hhmanager_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_201',['HHManager< Backend::MC, Device::CPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]] ]; diff --git a/master/search/all_8.js b/master/search/all_8.js index c9ccfcc131..1cad0ff5e2 100644 --- a/master/search/all_8.js +++ b/master/search/all_8.js @@ -1,34 +1,34 @@ var searchData= [ - ['index_2eh_201',['index.h',['../index_8h.html',1,'']]], - ['index2d_202',['Index2D',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common::Index2D< IndexT, Tag >'],['../communicator__grid_8h.html#af5204c68f1436b89ad06c653b710432e',1,'dlaf::comm::Index2D()'],['../classdlaf_1_1common_1_1_index2_d.html#ae959b1b0c405d3480f9d4960946b5b2e',1,'dlaf::common::Index2D::Index2D(const std::array< IndexT, 2 > &coords) noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#a940432586f1b501b8dda46355999335d',1,'dlaf::common::Index2D::Index2D() noexcept']]], - ['index2d_2eh_203',['index2d.h',['../index2d_8h.html',1,'']]], - ['index2d_3c_20indext_5fmpi_2c_20tag_5fmpi_20_3e_204',['Index2D< IndexT_MPI, TAG_MPI >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20sizetype_2c_20matrix_3a_3aglobalelement_5ftag_20_3e_205',['Index2D< SizeType, matrix::GlobalElement_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20sizetype_2c_20matrix_3a_3aglobaltile_5ftag_20_3e_206',['Index2D< SizeType, matrix::GlobalTile_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20sizetype_2c_20matrix_3a_3alocaltile_5ftag_20_3e_207',['Index2D< SizeType, matrix::LocalTile_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20sizetype_2c_20matrix_3a_3atileelement_5ftag_20_3e_208',['Index2D< SizeType, matrix::TileElement_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['indext_5fmpi_209',['IndexT_MPI',['../communicator_8h.html#a2b48f62607e5e128dff19ceac22093e9',1,'dlaf::comm']]], - ['init_2eh_210',['init.h',['../communication_2init_8h.html',1,'(Global Namespace)'],['../init_8h.html',1,'(Global Namespace)']]], - ['initialize_211',['initialize',['../init_8h.html#a94175c61baedbcc0bf6bd53e0f57380f',1,'dlaf::initialize(int argc, const char *const argv[], const configuration &user_cfg={})'],['../init_8h.html#a100757e13ff576c8a93dd7aed3ca7ec4',1,'dlaf::initialize(const pika::program_options::variables_map &vm, const configuration &user_cfg={})']]], - ['initresourcepartitionerhandler_212',['initResourcePartitionerHandler',['../init_8h.html#a3685b1a9ef2da35287ae3b03ea3b62e6',1,'dlaf']]], - ['internal_5f_213',['internal_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#adf84f8c3806212547305082bf21ac006',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['is_5fcontiguous_214',['is_contiguous',['../structdlaf_1_1common_1_1_data_descriptor.html#ad73cebe7217dd9f3ad68190dc09bce9e',1,'dlaf::common::DataDescriptor']]], - ['is_5fcoord_215',['is_coord',['../structdlaf_1_1common_1_1internal_1_1is__coord.html',1,'dlaf::common::internal']]], - ['is_5fcoord_3c_20index2d_3c_20t_2c_20tag_20_3e_20_3e_216',['is_coord< Index2D< T, Tag > >',['../structdlaf_1_1common_1_1internal_1_1is__coord_3_01_index2_d_3_01_t_00_01_tag_01_4_01_4.html',1,'dlaf::common::internal']]], - ['is_5fcoord_3c_20size2d_3c_20t_2c_20tag_20_3e_20_3e_217',['is_coord< Size2D< T, Tag > >',['../structdlaf_1_1common_1_1internal_1_1is__coord_3_01_size2_d_3_01_t_00_01_tag_01_4_01_4.html',1,'dlaf::common::internal']]], - ['is_5fdata_218',['is_data',['../structdlaf_1_1common_1_1is__data.html',1,'dlaf::common']]], - ['isempty_219',['isEmpty',['../classdlaf_1_1common_1_1_size2_d.html#afe5870841ddc785db5d4c81cc22aeb93',1,'dlaf::common::Size2D']]], - ['isexternal_220',['isExternal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa77cf35507fc6a6291e9197996034db8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['isfalse_221',['IsFalse',['../structdlaf_1_1internal_1_1_is_false.html',1,'dlaf::internal']]], - ['isfloatingpointorcomplex_222',['IsFloatingPointOrComplex',['../structdlaf_1_1internal_1_1_is_floating_point_or_complex.html',1,'dlaf::internal']]], - ['isfloatingpointorcomplex_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_223',['IsFloatingPointOrComplex< std::complex< T > >',['../structdlaf_1_1internal_1_1_is_floating_point_or_complex_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::internal']]], - ['isin_224',['isIn',['../classdlaf_1_1common_1_1_index2_d.html#a5f174cc234af20d79923b31d7fa90b19',1,'dlaf::common::Index2D']]], - ['isindexinrange_225',['isIndexInRange',['../classdlaf_1_1common_1_1_iterable_range2_d.html#a833f81c7ed8c08d0e995d9c97e8aa935',1,'dlaf::common::IterableRange2D']]], - ['isinoron_226',['isInOrOn',['../classdlaf_1_1common_1_1_index2_d.html#ab230f7ada23f15ffd3b3d0a3921f1685',1,'dlaf::common::Index2D']]], - ['isvalid_227',['isValid',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a380f1ac07da35bc014bcbfff908eb0a9',1,'dlaf::common::internal::basic_coords']]], - ['iterablerange2d_228',['IterableRange2D',['../classdlaf_1_1common_1_1_iterable_range2_d.html',1,'dlaf::common']]], - ['iterate_5frange2d_229',['iterate_range2d',['../range2d_8h.html#adcdcf40345b7f03cf52e871a2420057b',1,'dlaf::common']]], - ['iteratorlocal_230',['iteratorLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a79626e0d68d508aca69d1114cedc38a7',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::iteratorLocal()'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7c25686d3fb7e47b627af360592168ad',1,'dlaf::matrix::internal::View::iteratorLocal()']]], - ['iteratorrange2d_231',['IteratorRange2D',['../structdlaf_1_1common_1_1_iterator_range2_d.html',1,'dlaf::common']]] + ['index_2eh_202',['index.h',['../index_8h.html',1,'']]], + ['index2d_203',['Index2D',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common::Index2D< IndexT, Tag >'],['../communicator__grid_8h.html#af5204c68f1436b89ad06c653b710432e',1,'dlaf::comm::Index2D()'],['../classdlaf_1_1common_1_1_index2_d.html#ae959b1b0c405d3480f9d4960946b5b2e',1,'dlaf::common::Index2D::Index2D(const std::array< IndexT, 2 > &coords) noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#a940432586f1b501b8dda46355999335d',1,'dlaf::common::Index2D::Index2D() noexcept']]], + ['index2d_2eh_204',['index2d.h',['../index2d_8h.html',1,'']]], + ['index2d_3c_20indext_5fmpi_2c_20tag_5fmpi_20_3e_205',['Index2D< IndexT_MPI, TAG_MPI >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20sizetype_2c_20matrix_3a_3aglobalelement_5ftag_20_3e_206',['Index2D< SizeType, matrix::GlobalElement_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20sizetype_2c_20matrix_3a_3aglobaltile_5ftag_20_3e_207',['Index2D< SizeType, matrix::GlobalTile_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20sizetype_2c_20matrix_3a_3alocaltile_5ftag_20_3e_208',['Index2D< SizeType, matrix::LocalTile_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20sizetype_2c_20matrix_3a_3atileelement_5ftag_20_3e_209',['Index2D< SizeType, matrix::TileElement_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['indext_5fmpi_210',['IndexT_MPI',['../communicator_8h.html#a2b48f62607e5e128dff19ceac22093e9',1,'dlaf::comm']]], + ['init_2eh_211',['init.h',['../communication_2init_8h.html',1,'(Global Namespace)'],['../init_8h.html',1,'(Global Namespace)']]], + ['initialize_212',['initialize',['../init_8h.html#a94175c61baedbcc0bf6bd53e0f57380f',1,'dlaf::initialize(int argc, const char *const argv[], const configuration &user_cfg={})'],['../init_8h.html#a100757e13ff576c8a93dd7aed3ca7ec4',1,'dlaf::initialize(const pika::program_options::variables_map &vm, const configuration &user_cfg={})']]], + ['initresourcepartitionerhandler_213',['initResourcePartitionerHandler',['../init_8h.html#a3685b1a9ef2da35287ae3b03ea3b62e6',1,'dlaf']]], + ['internal_5f_214',['internal_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#adf84f8c3806212547305082bf21ac006',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['is_5fcontiguous_215',['is_contiguous',['../structdlaf_1_1common_1_1_data_descriptor.html#ad73cebe7217dd9f3ad68190dc09bce9e',1,'dlaf::common::DataDescriptor']]], + ['is_5fcoord_216',['is_coord',['../structdlaf_1_1common_1_1internal_1_1is__coord.html',1,'dlaf::common::internal']]], + ['is_5fcoord_3c_20index2d_3c_20t_2c_20tag_20_3e_20_3e_217',['is_coord< Index2D< T, Tag > >',['../structdlaf_1_1common_1_1internal_1_1is__coord_3_01_index2_d_3_01_t_00_01_tag_01_4_01_4.html',1,'dlaf::common::internal']]], + ['is_5fcoord_3c_20size2d_3c_20t_2c_20tag_20_3e_20_3e_218',['is_coord< Size2D< T, Tag > >',['../structdlaf_1_1common_1_1internal_1_1is__coord_3_01_size2_d_3_01_t_00_01_tag_01_4_01_4.html',1,'dlaf::common::internal']]], + ['is_5fdata_219',['is_data',['../structdlaf_1_1common_1_1is__data.html',1,'dlaf::common']]], + ['isempty_220',['isEmpty',['../classdlaf_1_1common_1_1_size2_d.html#afe5870841ddc785db5d4c81cc22aeb93',1,'dlaf::common::Size2D']]], + ['isexternal_221',['isExternal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa77cf35507fc6a6291e9197996034db8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['isfalse_222',['IsFalse',['../structdlaf_1_1internal_1_1_is_false.html',1,'dlaf::internal']]], + ['isfloatingpointorcomplex_223',['IsFloatingPointOrComplex',['../structdlaf_1_1internal_1_1_is_floating_point_or_complex.html',1,'dlaf::internal']]], + ['isfloatingpointorcomplex_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_224',['IsFloatingPointOrComplex< std::complex< T > >',['../structdlaf_1_1internal_1_1_is_floating_point_or_complex_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::internal']]], + ['isin_225',['isIn',['../classdlaf_1_1common_1_1_index2_d.html#a5f174cc234af20d79923b31d7fa90b19',1,'dlaf::common::Index2D']]], + ['isindexinrange_226',['isIndexInRange',['../classdlaf_1_1common_1_1_iterable_range2_d.html#a833f81c7ed8c08d0e995d9c97e8aa935',1,'dlaf::common::IterableRange2D']]], + ['isinoron_227',['isInOrOn',['../classdlaf_1_1common_1_1_index2_d.html#ab230f7ada23f15ffd3b3d0a3921f1685',1,'dlaf::common::Index2D']]], + ['isvalid_228',['isValid',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a380f1ac07da35bc014bcbfff908eb0a9',1,'dlaf::common::internal::basic_coords']]], + ['iterablerange2d_229',['IterableRange2D',['../classdlaf_1_1common_1_1_iterable_range2_d.html',1,'dlaf::common']]], + ['iterate_5frange2d_230',['iterate_range2d',['../range2d_8h.html#adcdcf40345b7f03cf52e871a2420057b',1,'dlaf::common']]], + ['iteratorlocal_231',['iteratorLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a79626e0d68d508aca69d1114cedc38a7',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::iteratorLocal()'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7c25686d3fb7e47b627af360592168ad',1,'dlaf::matrix::internal::View::iteratorLocal()']]], + ['iteratorrange2d_232',['IteratorRange2D',['../structdlaf_1_1common_1_1_iterator_range2_d.html',1,'dlaf::common']]] ]; diff --git a/master/search/all_9.js b/master/search/all_9.js index 24868e3d80..0215e5606b 100644 --- a/master/search/all_9.js +++ b/master/search/all_9.js @@ -1,4 +1,4 @@ var searchData= [ - ['kernels_2eh_232',['kernels.h',['../communication_2kernels_8h.html',1,'']]] + ['kernels_2eh_233',['kernels.h',['../communication_2kernels_8h.html',1,'']]] ]; diff --git a/master/search/all_a.js b/master/search/all_a.js index 16075dbf41..3b05723af7 100644 --- a/master/search/all_a.js +++ b/master/search/all_a.js @@ -1,19 +1,19 @@ var searchData= [ - ['lacpy_233',['lacpy',['../util__lapack_8h.html#a2a62e0ec6768b42ae19efff5ea5e4331',1,'lapack::lacpy()'],['../lapack_2tile_8h.html#a1bbf24144aebc55d9d50bb3b53ae38f7',1,'dlaf::tile::lacpy(TileElementSize region, TileElementIndex in_idx, const Tile< const T, Device::CPU > &in, TileElementIndex out_idx, const Tile< T, Device::CPU > &out)'],['../lapack_2tile_8h.html#adba78fac5535c410d8bd2e6c53ff8c7d',1,'dlaf::tile::lacpy(const Tile< const T, Device::CPU > &a, const Tile< T, Device::CPU > &b)']]], - ['lange_234',['lange',['../lapack_2tile_8h.html#a839d86dad6ce772207964d76b3ad07fa',1,'dlaf::tile']]], - ['lantr_235',['lantr',['../lapack_2tile_8h.html#a22be986552edcbdc4a3785a72620d0ca',1,'dlaf::tile']]], - ['laset_236',['laset',['../lapack_2tile_8h.html#a19439efac55cc846b684ed2c6f78eb33',1,'dlaf::tile::laset()'],['../util__lapack_8h.html#a653b4d466d700c7d6d3befb51c8ca9d8',1,'lapack::laset()']]], - ['layout_5finfo_2eh_237',['layout_info.h',['../layout__info_8h.html',1,'']]], - ['layoutinfo_238',['LayoutInfo',['../classdlaf_1_1matrix_1_1_layout_info.html',1,'dlaf::matrix::LayoutInfo'],['../classdlaf_1_1matrix_1_1_layout_info.html#a8c41196b99f84410426dfdd62609b982',1,'dlaf::matrix::LayoutInfo::LayoutInfo()']]], - ['ld_239',['ld',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a44282c10a5112a78c95c557538f13a74',1,'dlaf::matrix::Tile< const T, D >']]], - ['linearindex_240',['linearIndex',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a1d863b84e40d0567f731413ff788a1a6',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['local_5fmatrix_241',['local_matrix',['../util__matrix_8h.html#a801d4c1863ecd48a5758741b63781921',1,'dlaf::matrix']]], - ['localelementdistancefromglobaltile_242',['localElementDistanceFromGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#a532cc910ad683b4bf2154e5b01540e61',1,'dlaf::matrix::Distribution::localElementDistanceFromGlobalTile(GlobalTileIndex begin, GlobalTileIndex end) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ac0eef1aae43a7fc895cd280cac21eaa7',1,'dlaf::matrix::Distribution::localElementDistanceFromGlobalTile(SizeType i_begin, SizeType i_end) const noexcept']]], - ['localelementdistancefromlocaltile_243',['localElementDistanceFromLocalTile',['../classdlaf_1_1matrix_1_1_distribution.html#ab590472ea7c2132b30b725954f8ea3c0',1,'dlaf::matrix::Distribution::localElementDistanceFromLocalTile(LocalTileIndex begin, LocalTileIndex end) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#a757d638249f12dd13ffbaa95a9aba21a',1,'dlaf::matrix::Distribution::localElementDistanceFromLocalTile(SizeType i_loc_begin, SizeType i_loc_end) const noexcept']]], - ['localnrtiles_244',['localNrTiles',['../classdlaf_1_1matrix_1_1_distribution.html#a6e8c79e3f901845182c4e7e32de5eb34',1,'dlaf::matrix::Distribution']]], - ['localtilefromglobalelement_245',['localTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#ae812ba2b5ba6b753d733abb48acc5d72',1,'dlaf::matrix::Distribution']]], - ['localtilefromglobaltile_246',['localTileFromGlobalTile',['../util__distribution_8h.html#a235a8d1c90d534e32f93f2ce0e364700',1,'dlaf::util::matrix::localTileFromGlobalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#abdc93bce29920c2dff64235391cf3811',1,'dlaf::matrix::Distribution::localTileFromGlobalTile(SizeType global_tile) const noexcept']]], - ['localtileindex_247',['localTileIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a77f2413076c87fee45f6afb6e6e1baae',1,'dlaf::matrix::Distribution']]], - ['localtilelinearindex_248',['localTileLinearIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a22259431b76866f2f6acfe1a47ab3ace',1,'dlaf::matrix::Distribution']]] + ['lacpy_234',['lacpy',['../util__lapack_8h.html#a2a62e0ec6768b42ae19efff5ea5e4331',1,'lapack::lacpy()'],['../lapack_2tile_8h.html#a1bbf24144aebc55d9d50bb3b53ae38f7',1,'dlaf::tile::lacpy(TileElementSize region, TileElementIndex in_idx, const Tile< const T, Device::CPU > &in, TileElementIndex out_idx, const Tile< T, Device::CPU > &out)'],['../lapack_2tile_8h.html#adba78fac5535c410d8bd2e6c53ff8c7d',1,'dlaf::tile::lacpy(const Tile< const T, Device::CPU > &a, const Tile< T, Device::CPU > &b)']]], + ['lange_235',['lange',['../lapack_2tile_8h.html#a839d86dad6ce772207964d76b3ad07fa',1,'dlaf::tile']]], + ['lantr_236',['lantr',['../lapack_2tile_8h.html#a22be986552edcbdc4a3785a72620d0ca',1,'dlaf::tile']]], + ['laset_237',['laset',['../lapack_2tile_8h.html#a19439efac55cc846b684ed2c6f78eb33',1,'dlaf::tile::laset()'],['../util__lapack_8h.html#a653b4d466d700c7d6d3befb51c8ca9d8',1,'lapack::laset()']]], + ['layout_5finfo_2eh_238',['layout_info.h',['../layout__info_8h.html',1,'']]], + ['layoutinfo_239',['LayoutInfo',['../classdlaf_1_1matrix_1_1_layout_info.html',1,'dlaf::matrix::LayoutInfo'],['../classdlaf_1_1matrix_1_1_layout_info.html#a8c41196b99f84410426dfdd62609b982',1,'dlaf::matrix::LayoutInfo::LayoutInfo()']]], + ['ld_240',['ld',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a44282c10a5112a78c95c557538f13a74',1,'dlaf::matrix::Tile< const T, D >']]], + ['linearindex_241',['linearIndex',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a1d863b84e40d0567f731413ff788a1a6',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['local_5fmatrix_242',['local_matrix',['../util__matrix_8h.html#a801d4c1863ecd48a5758741b63781921',1,'dlaf::matrix']]], + ['localelementdistancefromglobaltile_243',['localElementDistanceFromGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#a532cc910ad683b4bf2154e5b01540e61',1,'dlaf::matrix::Distribution::localElementDistanceFromGlobalTile(GlobalTileIndex begin, GlobalTileIndex end) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ac0eef1aae43a7fc895cd280cac21eaa7',1,'dlaf::matrix::Distribution::localElementDistanceFromGlobalTile(SizeType i_begin, SizeType i_end) const noexcept']]], + ['localelementdistancefromlocaltile_244',['localElementDistanceFromLocalTile',['../classdlaf_1_1matrix_1_1_distribution.html#ab590472ea7c2132b30b725954f8ea3c0',1,'dlaf::matrix::Distribution::localElementDistanceFromLocalTile(LocalTileIndex begin, LocalTileIndex end) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#a757d638249f12dd13ffbaa95a9aba21a',1,'dlaf::matrix::Distribution::localElementDistanceFromLocalTile(SizeType i_loc_begin, SizeType i_loc_end) const noexcept']]], + ['localnrtiles_245',['localNrTiles',['../classdlaf_1_1matrix_1_1_distribution.html#a6e8c79e3f901845182c4e7e32de5eb34',1,'dlaf::matrix::Distribution']]], + ['localtilefromglobalelement_246',['localTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#ae812ba2b5ba6b753d733abb48acc5d72',1,'dlaf::matrix::Distribution']]], + ['localtilefromglobaltile_247',['localTileFromGlobalTile',['../util__distribution_8h.html#a235a8d1c90d534e32f93f2ce0e364700',1,'dlaf::util::matrix::localTileFromGlobalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#abdc93bce29920c2dff64235391cf3811',1,'dlaf::matrix::Distribution::localTileFromGlobalTile(SizeType global_tile) const noexcept']]], + ['localtileindex_248',['localTileIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a77f2413076c87fee45f6afb6e6e1baae',1,'dlaf::matrix::Distribution']]], + ['localtilelinearindex_249',['localTileLinearIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a22259431b76866f2f6acfe1a47ab3ace',1,'dlaf::matrix::Distribution']]] ]; diff --git a/master/search/all_b.js b/master/search/all_b.js index 97eab69e49..812b891d72 100644 --- a/master/search/all_b.js +++ b/master/search/all_b.js @@ -1,41 +1,41 @@ var searchData= [ - ['make_5fcommunicator_5fmanaged_249',['make_communicator_managed',['../communicator_8h.html#a8b7bf4fa5dfb408439c9c93ce3ebb216',1,'dlaf::comm::make_communicator_managed()'],['../classdlaf_1_1comm_1_1_communicator.html#ada042e7529d465519d9fb98e004d6a55',1,'dlaf::comm::Communicator::make_communicator_managed()']]], - ['make_5fcontiguous_250',['make_contiguous',['../data__descriptor_8h.html#a96a67d2223bad7ea28fcca7041877560',1,'dlaf::common']]], - ['make_5fdata_251',['make_data',['../data_8h.html#aa3aa1962f3a8382cbf457db22c8ca48f',1,'dlaf::common::make_data(T &&data, Ts &&... args) noexcept'],['../data_8h.html#ae8a4cf6d87f0a72e5daad8a35478f051',1,'dlaf::common::make_data(Data &&data) noexcept']]], - ['make_5fmessage_252',['make_message',['../message_8h.html#a256282d96e42c4ff2e6353ae28263422',1,'dlaf::comm']]], - ['matrix_253',['Matrix',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix::Matrix< T, D >'],['../classdlaf_1_1matrix_1_1_matrix.html#a5d2827ca2dbad4d6e6482e4c1e963472',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution, const LayoutInfo &layout, ElementType *ptr) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#ad5f429b579c79194b453206f24d7de49',1,'dlaf::matrix::Matrix::Matrix(const LayoutInfo &layout, ElementType *ptr)'],['../classdlaf_1_1matrix_1_1_matrix.html#a1f0c98f29fa14cccef9f5378f6ee20b1',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution, const LayoutInfo &layout) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#a73b38df3414aa5272abd3b6e87d67e37',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution)'],['../classdlaf_1_1matrix_1_1_matrix.html#a65b04ca6f55cf213c52cda2f3a532023',1,'dlaf::matrix::Matrix::Matrix(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm)'],['../classdlaf_1_1matrix_1_1_matrix.html#aa0614a38addeb9aaf83d268f1092b779',1,'dlaf::matrix::Matrix::Matrix(const LocalElementSize &size, const TileElementSize &block_size)']]], - ['matrix_2eh_254',['matrix.h',['../matrix_8h.html',1,'']]], - ['matrix_3c_20const_20t_2c_20d_20_3e_255',['Matrix< const T, D >',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html',1,'dlaf::matrix']]], - ['matrix_3c_20const_20t_2c_20source_20_3e_256',['Matrix< const T, Source >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20const_20t_2c_20sourcetarget_20_3e_257',['Matrix< const T, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20t_2c_20source_20_3e_258',['Matrix< T, Source >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20t_2c_20sourcetarget_20_3e_259',['Matrix< T, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20t_2c_20target_20_3e_260',['Matrix< T, Target >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_5fbase_2eh_261',['matrix_base.h',['../matrix__base_8h.html',1,'']]], - ['matrix_5fmirror_2eh_262',['matrix_mirror.h',['../matrix__mirror_8h.html',1,'']]], - ['matrixbase_263',['MatrixBase',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html',1,'dlaf::matrix::internal']]], - ['matrixmirror_264',['MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror.html',1,'dlaf::matrix::MatrixMirror< T, Target, Source >'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#ab4b4d7a17b5f0ae45d301ef415f41b29',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a66b7ac6e591f693cd7bb8119955055fc',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#ac9dfb2578a5c94b8cf6bf0d5b3186979',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#a355ab06273bb75ba0fb40119829d2aa0',1,'dlaf::matrix::MatrixMirror::MatrixMirror()']]], - ['matrixmirror_3c_20const_20t_2c_20sourcetarget_2c_20sourcetarget_20_3e_265',['MatrixMirror< const T, SourceTarget, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html',1,'dlaf::matrix']]], - ['matrixmirror_3c_20const_20t_2c_20target_2c_20source_20_3e_266',['MatrixMirror< const T, Target, Source >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html',1,'dlaf::matrix']]], - ['matrixmirror_3c_20t_2c_20sourcetarget_2c_20sourcetarget_20_3e_267',['MatrixMirror< T, SourceTarget, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html',1,'dlaf::matrix']]], - ['memory_5fchunk_2eh_268',['memory_chunk.h',['../memory__chunk_8h.html',1,'']]], - ['memory_5fview_2eh_269',['memory_view.h',['../memory__view_8h.html',1,'']]], - ['memorychunk_270',['MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html',1,'dlaf::memory::MemoryChunk< T, D >'],['../classdlaf_1_1memory_1_1_memory_chunk.html#ab7a334b89c214fe25ffde4bf0fd4e802',1,'dlaf::memory::MemoryChunk::MemoryChunk(MemoryChunk &&rhs) noexcept'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a8855d0d0e0c35c83394ce82cd14ea36d',1,'dlaf::memory::MemoryChunk::MemoryChunk(T *ptr, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a53cffddd0ddfd2e1e0db839157a58913',1,'dlaf::memory::MemoryChunk::MemoryChunk(SizeType size)'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a4059fde641a91f8cac05721fb54c749b',1,'dlaf::memory::MemoryChunk::MemoryChunk()']]], - ['memoryview_271',['MemoryView',['../classdlaf_1_1memory_1_1_memory_view.html',1,'dlaf::memory::MemoryView< T, D >'],['../classdlaf_1_1memory_1_1_memory_view.html#aa0752bb38aa279bd6260e2d440b8ca7d',1,'dlaf::memory::MemoryView::MemoryView()'],['../classdlaf_1_1memory_1_1_memory_view.html#a0650b61938741469c7583d671240619d',1,'dlaf::memory::MemoryView::MemoryView(SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#a0835f719277ae4f7b97c68d097a07071',1,'dlaf::memory::MemoryView::MemoryView(T *ptr, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#a76c5e9d28626abb2991dd6ad1ce460c7',1,'dlaf::memory::MemoryView::MemoryView(const MemoryView &memory_view, SizeType offset, SizeType size)']]], - ['memoryview_3c_20t_2c_20device_3a_3acpu_20_3e_272',['MemoryView< T, Device::CPU >',['../classdlaf_1_1memory_1_1_memory_view.html',1,'dlaf::memory']]], - ['message_273',['Message',['../classdlaf_1_1comm_1_1_message.html',1,'dlaf::comm::Message< Data >'],['../classdlaf_1_1comm_1_1_message.html#aa0a15382490f3f628d14f734aeddf285',1,'dlaf::comm::Message::Message()']]], - ['message_2eh_274',['message.h',['../message_8h.html',1,'']]], - ['minmemsize_275',['minMemSize',['../classdlaf_1_1matrix_1_1_layout_info.html#ad51fa263d160dce4f4db2943720d2506',1,'dlaf::matrix::LayoutInfo']]], - ['mintilememsize_276',['minTileMemSize',['../classdlaf_1_1matrix_1_1_layout_info.html#a90af3a4b4a456d3107dec3eb4c207c29',1,'dlaf::matrix::LayoutInfo::minTileMemSize(const LocalTileIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_layout_info.html#aa13f1acc6c4df1d0e0c11204d3c81ba0',1,'dlaf::matrix::LayoutInfo::minTileMemSize(const TileElementSize &tile_size) const noexcept']]], - ['movenonconsttile_277',['moveNonConstTile',['../structdlaf_1_1internal_1_1move_non_const_tile.html',1,'dlaf::internal']]], - ['mpi_5fcall_278',['mpi_call',['../communication_2error_8h.html#aa640fad0edb091d239f68708fbe00103',1,'dlaf::internal']]], - ['mpi_5fdatatype_279',['mpi_datatype',['../structdlaf_1_1comm_1_1mpi__datatype.html',1,'dlaf::comm']]], - ['mpi_5fdatatype_3c_20const_20t_20_3e_280',['mpi_datatype< const T >',['../structdlaf_1_1comm_1_1mpi__datatype_3_01const_01_t_01_4.html',1,'dlaf::comm']]], - ['mpi_5finit_281',['mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html',1,'dlaf::comm::mpi_init'],['../structdlaf_1_1comm_1_1mpi__init.html#a415e55c02e205e6f782ff7000755c842',1,'dlaf::comm::mpi_init::mpi_init()']]], - ['mpi_5ftype_282',['mpi_type',['../classdlaf_1_1comm_1_1_message.html#ac9fa892f3513f9b60ad178d1f788f0af',1,'dlaf::comm::Message']]], - ['mpicallhelper_283',['MPICallHelper',['../structdlaf_1_1comm_1_1internal_1_1_m_p_i_call_helper.html',1,'dlaf::comm::internal']]], - ['mul_284',['mul',['../util__math_8h.html#a77a106250e33d119c5415546f6b74544',1,'dlaf::util::size_t::mul()'],['../util__math_8h.html#a090c5b89334c2c21574610d8f0881b67',1,'dlaf::util::ptrdiff_t::mul()']]], - ['multipliable_285',['multipliable',['../util__matrix_8h.html#a14086310f58c9ace5ceb41e040277c86',1,'dlaf::matrix']]], - ['multipliable_5fsizes_286',['multipliable_sizes',['../util__matrix_8h.html#a93b53e9c0e942a5caa96a24242864400',1,'dlaf::matrix']]] + ['make_5fcommunicator_5fmanaged_250',['make_communicator_managed',['../communicator_8h.html#a8b7bf4fa5dfb408439c9c93ce3ebb216',1,'dlaf::comm::make_communicator_managed()'],['../classdlaf_1_1comm_1_1_communicator.html#ada042e7529d465519d9fb98e004d6a55',1,'dlaf::comm::Communicator::make_communicator_managed()']]], + ['make_5fcontiguous_251',['make_contiguous',['../data__descriptor_8h.html#a96a67d2223bad7ea28fcca7041877560',1,'dlaf::common']]], + ['make_5fdata_252',['make_data',['../data_8h.html#aa3aa1962f3a8382cbf457db22c8ca48f',1,'dlaf::common::make_data(T &&data, Ts &&... args) noexcept'],['../data_8h.html#ae8a4cf6d87f0a72e5daad8a35478f051',1,'dlaf::common::make_data(Data &&data) noexcept']]], + ['make_5fmessage_253',['make_message',['../message_8h.html#a256282d96e42c4ff2e6353ae28263422',1,'dlaf::comm']]], + ['matrix_254',['Matrix',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix::Matrix< T, D >'],['../classdlaf_1_1matrix_1_1_matrix.html#a5d2827ca2dbad4d6e6482e4c1e963472',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution, const LayoutInfo &layout, ElementType *ptr) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#ad5f429b579c79194b453206f24d7de49',1,'dlaf::matrix::Matrix::Matrix(const LayoutInfo &layout, ElementType *ptr)'],['../classdlaf_1_1matrix_1_1_matrix.html#a1f0c98f29fa14cccef9f5378f6ee20b1',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution, const LayoutInfo &layout) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#a73b38df3414aa5272abd3b6e87d67e37',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution)'],['../classdlaf_1_1matrix_1_1_matrix.html#a65b04ca6f55cf213c52cda2f3a532023',1,'dlaf::matrix::Matrix::Matrix(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm)'],['../classdlaf_1_1matrix_1_1_matrix.html#aa0614a38addeb9aaf83d268f1092b779',1,'dlaf::matrix::Matrix::Matrix(const LocalElementSize &size, const TileElementSize &block_size)']]], + ['matrix_2eh_255',['matrix.h',['../matrix_8h.html',1,'']]], + ['matrix_3c_20const_20t_2c_20d_20_3e_256',['Matrix< const T, D >',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html',1,'dlaf::matrix']]], + ['matrix_3c_20const_20t_2c_20source_20_3e_257',['Matrix< const T, Source >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20const_20t_2c_20sourcetarget_20_3e_258',['Matrix< const T, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20t_2c_20source_20_3e_259',['Matrix< T, Source >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20t_2c_20sourcetarget_20_3e_260',['Matrix< T, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20t_2c_20target_20_3e_261',['Matrix< T, Target >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_5fbase_2eh_262',['matrix_base.h',['../matrix__base_8h.html',1,'']]], + ['matrix_5fmirror_2eh_263',['matrix_mirror.h',['../matrix__mirror_8h.html',1,'']]], + ['matrixbase_264',['MatrixBase',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html',1,'dlaf::matrix::internal']]], + ['matrixmirror_265',['MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror.html',1,'dlaf::matrix::MatrixMirror< T, Target, Source >'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#ab4b4d7a17b5f0ae45d301ef415f41b29',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a66b7ac6e591f693cd7bb8119955055fc',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#ac9dfb2578a5c94b8cf6bf0d5b3186979',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#a355ab06273bb75ba0fb40119829d2aa0',1,'dlaf::matrix::MatrixMirror::MatrixMirror()']]], + ['matrixmirror_3c_20const_20t_2c_20sourcetarget_2c_20sourcetarget_20_3e_266',['MatrixMirror< const T, SourceTarget, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html',1,'dlaf::matrix']]], + ['matrixmirror_3c_20const_20t_2c_20target_2c_20source_20_3e_267',['MatrixMirror< const T, Target, Source >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html',1,'dlaf::matrix']]], + ['matrixmirror_3c_20t_2c_20sourcetarget_2c_20sourcetarget_20_3e_268',['MatrixMirror< T, SourceTarget, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html',1,'dlaf::matrix']]], + ['memory_5fchunk_2eh_269',['memory_chunk.h',['../memory__chunk_8h.html',1,'']]], + ['memory_5fview_2eh_270',['memory_view.h',['../memory__view_8h.html',1,'']]], + ['memorychunk_271',['MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html',1,'dlaf::memory::MemoryChunk< T, D >'],['../classdlaf_1_1memory_1_1_memory_chunk.html#ab7a334b89c214fe25ffde4bf0fd4e802',1,'dlaf::memory::MemoryChunk::MemoryChunk(MemoryChunk &&rhs) noexcept'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a8855d0d0e0c35c83394ce82cd14ea36d',1,'dlaf::memory::MemoryChunk::MemoryChunk(T *ptr, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a53cffddd0ddfd2e1e0db839157a58913',1,'dlaf::memory::MemoryChunk::MemoryChunk(SizeType size)'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a4059fde641a91f8cac05721fb54c749b',1,'dlaf::memory::MemoryChunk::MemoryChunk()']]], + ['memoryview_272',['MemoryView',['../classdlaf_1_1memory_1_1_memory_view.html',1,'dlaf::memory::MemoryView< T, D >'],['../classdlaf_1_1memory_1_1_memory_view.html#aa0752bb38aa279bd6260e2d440b8ca7d',1,'dlaf::memory::MemoryView::MemoryView()'],['../classdlaf_1_1memory_1_1_memory_view.html#a0650b61938741469c7583d671240619d',1,'dlaf::memory::MemoryView::MemoryView(SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#a0835f719277ae4f7b97c68d097a07071',1,'dlaf::memory::MemoryView::MemoryView(T *ptr, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#a76c5e9d28626abb2991dd6ad1ce460c7',1,'dlaf::memory::MemoryView::MemoryView(const MemoryView &memory_view, SizeType offset, SizeType size)']]], + ['memoryview_3c_20t_2c_20device_3a_3acpu_20_3e_273',['MemoryView< T, Device::CPU >',['../classdlaf_1_1memory_1_1_memory_view.html',1,'dlaf::memory']]], + ['message_274',['Message',['../classdlaf_1_1comm_1_1_message.html',1,'dlaf::comm::Message< Data >'],['../classdlaf_1_1comm_1_1_message.html#aa0a15382490f3f628d14f734aeddf285',1,'dlaf::comm::Message::Message()']]], + ['message_2eh_275',['message.h',['../message_8h.html',1,'']]], + ['minmemsize_276',['minMemSize',['../classdlaf_1_1matrix_1_1_layout_info.html#ad51fa263d160dce4f4db2943720d2506',1,'dlaf::matrix::LayoutInfo']]], + ['mintilememsize_277',['minTileMemSize',['../classdlaf_1_1matrix_1_1_layout_info.html#a90af3a4b4a456d3107dec3eb4c207c29',1,'dlaf::matrix::LayoutInfo::minTileMemSize(const LocalTileIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_layout_info.html#aa13f1acc6c4df1d0e0c11204d3c81ba0',1,'dlaf::matrix::LayoutInfo::minTileMemSize(const TileElementSize &tile_size) const noexcept']]], + ['movenonconsttile_278',['moveNonConstTile',['../structdlaf_1_1internal_1_1move_non_const_tile.html',1,'dlaf::internal']]], + ['mpi_5fcall_279',['mpi_call',['../communication_2error_8h.html#aa640fad0edb091d239f68708fbe00103',1,'dlaf::internal']]], + ['mpi_5fdatatype_280',['mpi_datatype',['../structdlaf_1_1comm_1_1mpi__datatype.html',1,'dlaf::comm']]], + ['mpi_5fdatatype_3c_20const_20t_20_3e_281',['mpi_datatype< const T >',['../structdlaf_1_1comm_1_1mpi__datatype_3_01const_01_t_01_4.html',1,'dlaf::comm']]], + ['mpi_5finit_282',['mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html',1,'dlaf::comm::mpi_init'],['../structdlaf_1_1comm_1_1mpi__init.html#a415e55c02e205e6f782ff7000755c842',1,'dlaf::comm::mpi_init::mpi_init()']]], + ['mpi_5ftype_283',['mpi_type',['../classdlaf_1_1comm_1_1_message.html#ac9fa892f3513f9b60ad178d1f788f0af',1,'dlaf::comm::Message']]], + ['mpicallhelper_284',['MPICallHelper',['../structdlaf_1_1comm_1_1internal_1_1_m_p_i_call_helper.html',1,'dlaf::comm::internal']]], + ['mul_285',['mul',['../util__math_8h.html#a77a106250e33d119c5415546f6b74544',1,'dlaf::util::size_t::mul()'],['../util__math_8h.html#a090c5b89334c2c21574610d8f0881b67',1,'dlaf::util::ptrdiff_t::mul()']]], + ['multipliable_286',['multipliable',['../util__matrix_8h.html#a14086310f58c9ace5ceb41e040277c86',1,'dlaf::matrix']]], + ['multipliable_5fsizes_287',['multipliable_sizes',['../util__matrix_8h.html#a93b53e9c0e942a5caa96a24242864400',1,'dlaf::matrix']]] ]; diff --git a/master/search/all_c.js b/master/search/all_c.js index 596e3c1074..72f8b04d0a 100644 --- a/master/search/all_c.js +++ b/master/search/all_c.js @@ -1,14 +1,14 @@ var searchData= [ - ['nblocks_287',['nblocks',['../structdlaf_1_1common_1_1_data_descriptor.html#a583161cb3c61e3af88fe66552a123866',1,'dlaf::common::DataDescriptor']]], - ['nextlocaltilefromglobalelement_288',['nextLocalTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a559caf9af0f658507a6a8fe949e09a93',1,'dlaf::matrix::Distribution']]], - ['nextlocaltilefromglobaltile_289',['nextLocalTileFromGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#afcd76386ddd1319978242aaf54412825',1,'dlaf::matrix::Distribution::nextLocalTileFromGlobalTile()'],['../util__distribution_8h.html#ad6d38066f4c7c00e41fc2d8830efb77f',1,'dlaf::util::matrix::nextLocalTileFromGlobalTile()']]], - ['norm_290',['Norm',['../structdlaf_1_1auxiliary_1_1internal_1_1_norm.html',1,'dlaf::auxiliary::internal']]], - ['norm_291',['norm',['../norm_8h.html#a2d0adfb7be9cfe934f96e9dd1c38d839',1,'dlaf::auxiliary']]], - ['norm_2eh_292',['norm.h',['../norm_8h.html',1,'']]], - ['norm_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_293',['Norm< Backend::MC, Device::CPU, T >',['../structdlaf_1_1auxiliary_1_1internal_1_1_norm_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::auxiliary::internal']]], - ['nrtiles_294',['nrTiles',['../classdlaf_1_1matrix_1_1_distribution.html#acc85a62e6350bcac93a8e442d4603516',1,'dlaf::matrix::Distribution::nrTiles()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a068e58f7f6c2956f92e69fd7c8d8f68d',1,'dlaf::matrix::internal::MatrixBase::nrTiles()']]], - ['numpy_295',['numpy',['../structdlaf_1_1format_1_1numpy.html',1,'dlaf::format']]], - ['numpy_5fdatatype_296',['numpy_datatype',['../structdlaf_1_1matrix_1_1internal_1_1numpy__datatype.html',1,'dlaf::matrix::internal']]], - ['numpy_5fdatatype_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_297',['numpy_datatype< std::complex< T > >',['../structdlaf_1_1matrix_1_1internal_1_1numpy__datatype_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::matrix::internal']]] + ['nblocks_288',['nblocks',['../structdlaf_1_1common_1_1_data_descriptor.html#a583161cb3c61e3af88fe66552a123866',1,'dlaf::common::DataDescriptor']]], + ['nextlocaltilefromglobalelement_289',['nextLocalTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a559caf9af0f658507a6a8fe949e09a93',1,'dlaf::matrix::Distribution']]], + ['nextlocaltilefromglobaltile_290',['nextLocalTileFromGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#afcd76386ddd1319978242aaf54412825',1,'dlaf::matrix::Distribution::nextLocalTileFromGlobalTile()'],['../util__distribution_8h.html#ad6d38066f4c7c00e41fc2d8830efb77f',1,'dlaf::util::matrix::nextLocalTileFromGlobalTile()']]], + ['norm_291',['Norm',['../structdlaf_1_1auxiliary_1_1internal_1_1_norm.html',1,'dlaf::auxiliary::internal']]], + ['norm_292',['norm',['../norm_8h.html#a2d0adfb7be9cfe934f96e9dd1c38d839',1,'dlaf::auxiliary']]], + ['norm_2eh_293',['norm.h',['../norm_8h.html',1,'']]], + ['norm_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_294',['Norm< Backend::MC, Device::CPU, T >',['../structdlaf_1_1auxiliary_1_1internal_1_1_norm_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::auxiliary::internal']]], + ['nrtiles_295',['nrTiles',['../classdlaf_1_1matrix_1_1_distribution.html#acc85a62e6350bcac93a8e442d4603516',1,'dlaf::matrix::Distribution::nrTiles()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a068e58f7f6c2956f92e69fd7c8d8f68d',1,'dlaf::matrix::internal::MatrixBase::nrTiles()']]], + ['numpy_296',['numpy',['../structdlaf_1_1format_1_1numpy.html',1,'dlaf::format']]], + ['numpy_5fdatatype_297',['numpy_datatype',['../structdlaf_1_1matrix_1_1internal_1_1numpy__datatype.html',1,'dlaf::matrix::internal']]], + ['numpy_5fdatatype_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_298',['numpy_datatype< std::complex< T > >',['../structdlaf_1_1matrix_1_1internal_1_1numpy__datatype_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::matrix::internal']]] ]; diff --git a/master/search/all_d.js b/master/search/all_d.js index b2d4f3f8c7..4a75d4dcf9 100644 --- a/master/search/all_d.js +++ b/master/search/all_d.js @@ -1,17 +1,17 @@ var searchData= [ - ['offset_298',['offset',['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7229ff87452798d32b7195fe684fff7a',1,'dlaf::matrix::internal::View']]], - ['offsetelement_299',['offsetElement',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a398b31442f71b68e39a3ecbb2762203e',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::offsetElement()'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7164cccc46c63cc96661288809a37f5f',1,'dlaf::matrix::internal::View::offsetElement()']]], - ['operator_20bool_300',['operator bool',['../structdlaf_1_1common_1_1_buffer.html#ae57231c34ebf6f79da4d1e6e1c75cf0a',1,'dlaf::common::Buffer::operator bool()'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a8f6d65a5adc84620ed33a7de52bac602',1,'dlaf::comm::internal::type_handler::operator bool()']]], - ['operator_20mpi_5fcomm_301',['operator MPI_Comm',['../classdlaf_1_1comm_1_1_communicator.html#a5be809bbc2d038669e31a9de501a2070',1,'dlaf::comm::Communicator']]], - ['operator_20mpi_5fdatatype_302',['operator MPI_Datatype',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a2247361b0d7dd77039860c0cd21a13e3',1,'dlaf::comm::internal::type_handler']]], - ['operator_21_3d_303',['operator!=',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#abf2886043fbe929f1e1f0eff5e20935d',1,'dlaf::common::internal::basic_coords::operator!=()'],['../classdlaf_1_1common_1_1_size2_d.html#a1a6822fefe13303612ce0ddf44f260ea',1,'dlaf::common::Size2D::operator!=()'],['../classdlaf_1_1common_1_1_index2_d.html#ae2e28cf6eab4febd97d6c46d23c6096d',1,'dlaf::common::Index2D::operator!=()']]], - ['operator_26_304',['operator&',['../classdlaf_1_1comm_1_1_communicator.html#ace61947783c442bf684c2f035093acf2',1,'dlaf::comm::Communicator']]], - ['operator_28_29_305',['operator()',['../classdlaf_1_1memory_1_1_memory_chunk.html#a1720c41c3c3cbad2e198e7cedbfa84de',1,'dlaf::memory::MemoryChunk::operator()()'],['../classdlaf_1_1memory_1_1_memory_view.html#aad118548f59d6c8021eb4fbe816a1561',1,'dlaf::memory::MemoryView::operator()() const'],['../classdlaf_1_1memory_1_1_memory_view.html#a6f9c8123e686e2f2122ec988ca0b9725',1,'dlaf::memory::MemoryView::operator()(SizeType index) const'],['../classdlaf_1_1memory_1_1_memory_chunk.html#ac0325c028366b219597dde97814e9a7a',1,'dlaf::memory::MemoryChunk::operator()()'],['../classdlaf_1_1matrix_1_1_tile.html#a5641fc26b7eb759a31ddc24975393ce7',1,'dlaf::matrix::Tile::operator()()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#ac22706782f9864d6357df2f48f15ee74',1,'dlaf::matrix::Tile< const T, D >::operator()()'],['../classdlaf_1_1common_1_1_pipeline.html#a388b1fe53e84dde5b0f026e7b334902d',1,'dlaf::common::Pipeline::operator()()']]], - ['operator_2b_306',['operator+',['../index2d_8h.html#a2ed7f0c019788f600bc067956646bf75',1,'dlaf::common']]], - ['operator_3c_3c_307',['operator<<',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a9a65958ab367757d900bd588a405df77',1,'dlaf::common::internal::basic_coords::operator<<()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#ac215ee8142401d06aba72494c8aa8b9f',1,'dlaf::comm::CommunicatorGrid::operator<<()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a054e897882d7ab913c1b63af96e87fec',1,'dlaf::matrix::internal::MatrixBase::operator<<()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a45b5556596386e6bb46e0dc61b838811',1,'dlaf::matrix::Tile< const T, D >::operator<<()'],['../index2d_8h.html#a352cdd70886a2a2ace1abd246951c538',1,'dlaf::common::operator<<()']]], - ['operator_3d_308',['operator=',['../classdlaf_1_1memory_1_1_memory_chunk.html#a0c03f967f09e963519dae5e235da66e1',1,'dlaf::memory::MemoryChunk']]], - ['operator_3d_3d_309',['operator==',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a95e16def9c322ee963d123635c2c875b',1,'dlaf::common::internal::basic_coords::operator==()'],['../classdlaf_1_1common_1_1_size2_d.html#a83385feaa5e14b4291c6267bc68fbe51',1,'dlaf::common::Size2D::operator==()'],['../classdlaf_1_1common_1_1_index2_d.html#a4724fdad093b5d189c5f5e0b1ac5e305',1,'dlaf::common::Index2D::operator==()']]], - ['ordering_310',['Ordering',['../index2d_8h.html#a02419996f0ffc187079c310f0175ce76',1,'dlaf::common']]], - ['orthogonal_311',['orthogonal',['../index2d_8h.html#a9b3be798b052f32def2a280a80e2295c',1,'dlaf']]] + ['offset_299',['offset',['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7229ff87452798d32b7195fe684fff7a',1,'dlaf::matrix::internal::View']]], + ['offsetelement_300',['offsetElement',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a398b31442f71b68e39a3ecbb2762203e',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::offsetElement()'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7164cccc46c63cc96661288809a37f5f',1,'dlaf::matrix::internal::View::offsetElement()']]], + ['operator_20bool_301',['operator bool',['../structdlaf_1_1common_1_1_buffer.html#ae57231c34ebf6f79da4d1e6e1c75cf0a',1,'dlaf::common::Buffer::operator bool()'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a8f6d65a5adc84620ed33a7de52bac602',1,'dlaf::comm::internal::type_handler::operator bool()']]], + ['operator_20mpi_5fcomm_302',['operator MPI_Comm',['../classdlaf_1_1comm_1_1_communicator.html#a5be809bbc2d038669e31a9de501a2070',1,'dlaf::comm::Communicator']]], + ['operator_20mpi_5fdatatype_303',['operator MPI_Datatype',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a2247361b0d7dd77039860c0cd21a13e3',1,'dlaf::comm::internal::type_handler']]], + ['operator_21_3d_304',['operator!=',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#abf2886043fbe929f1e1f0eff5e20935d',1,'dlaf::common::internal::basic_coords::operator!=()'],['../classdlaf_1_1common_1_1_size2_d.html#a1a6822fefe13303612ce0ddf44f260ea',1,'dlaf::common::Size2D::operator!=()'],['../classdlaf_1_1common_1_1_index2_d.html#ae2e28cf6eab4febd97d6c46d23c6096d',1,'dlaf::common::Index2D::operator!=()']]], + ['operator_26_305',['operator&',['../classdlaf_1_1comm_1_1_communicator.html#ace61947783c442bf684c2f035093acf2',1,'dlaf::comm::Communicator']]], + ['operator_28_29_306',['operator()',['../classdlaf_1_1memory_1_1_memory_chunk.html#a1720c41c3c3cbad2e198e7cedbfa84de',1,'dlaf::memory::MemoryChunk::operator()()'],['../classdlaf_1_1memory_1_1_memory_view.html#aad118548f59d6c8021eb4fbe816a1561',1,'dlaf::memory::MemoryView::operator()() const'],['../classdlaf_1_1memory_1_1_memory_view.html#a6f9c8123e686e2f2122ec988ca0b9725',1,'dlaf::memory::MemoryView::operator()(SizeType index) const'],['../classdlaf_1_1memory_1_1_memory_chunk.html#ac0325c028366b219597dde97814e9a7a',1,'dlaf::memory::MemoryChunk::operator()()'],['../classdlaf_1_1matrix_1_1_tile.html#a5641fc26b7eb759a31ddc24975393ce7',1,'dlaf::matrix::Tile::operator()()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#ac22706782f9864d6357df2f48f15ee74',1,'dlaf::matrix::Tile< const T, D >::operator()()'],['../classdlaf_1_1common_1_1_pipeline.html#a388b1fe53e84dde5b0f026e7b334902d',1,'dlaf::common::Pipeline::operator()()']]], + ['operator_2b_307',['operator+',['../index2d_8h.html#a2ed7f0c019788f600bc067956646bf75',1,'dlaf::common']]], + ['operator_3c_3c_308',['operator<<',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a9a65958ab367757d900bd588a405df77',1,'dlaf::common::internal::basic_coords::operator<<()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#ac215ee8142401d06aba72494c8aa8b9f',1,'dlaf::comm::CommunicatorGrid::operator<<()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a054e897882d7ab913c1b63af96e87fec',1,'dlaf::matrix::internal::MatrixBase::operator<<()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a45b5556596386e6bb46e0dc61b838811',1,'dlaf::matrix::Tile< const T, D >::operator<<()'],['../index2d_8h.html#a352cdd70886a2a2ace1abd246951c538',1,'dlaf::common::operator<<()']]], + ['operator_3d_309',['operator=',['../classdlaf_1_1memory_1_1_memory_chunk.html#a0c03f967f09e963519dae5e235da66e1',1,'dlaf::memory::MemoryChunk']]], + ['operator_3d_3d_310',['operator==',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a95e16def9c322ee963d123635c2c875b',1,'dlaf::common::internal::basic_coords::operator==()'],['../classdlaf_1_1common_1_1_size2_d.html#a83385feaa5e14b4291c6267bc68fbe51',1,'dlaf::common::Size2D::operator==()'],['../classdlaf_1_1common_1_1_index2_d.html#a4724fdad093b5d189c5f5e0b1ac5e305',1,'dlaf::common::Index2D::operator==()']]], + ['ordering_311',['Ordering',['../index2d_8h.html#a02419996f0ffc187079c310f0175ce76',1,'dlaf::common']]], + ['orthogonal_312',['orthogonal',['../index2d_8h.html#a9b3be798b052f32def2a280a80e2295c',1,'dlaf']]] ]; diff --git a/master/search/all_e.js b/master/search/all_e.js index d95147e7a3..ac8ff9778e 100644 --- a/master/search/all_e.js +++ b/master/search/all_e.js @@ -1,31 +1,31 @@ var searchData= [ - ['p2p_2eh_312',['p2p.h',['../p2p_8h.html',1,'']]], - ['p2p_5fallsum_2eh_313',['p2p_allsum.h',['../p2p__allsum_8h.html',1,'']]], - ['panel_314',['Panel',['../structdlaf_1_1matrix_1_1_panel.html',1,'dlaf::matrix::Panel< axis, T, D, Storage >'],['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ad63a4d555cea1fdf2f6d5745125fe3cb',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::Panel()']]], - ['panel_2eh_315',['panel.h',['../panel_8h.html',1,'']]], - ['panel_3c_20axis_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ano_20_3e_316',['Panel< axis, const T, D, StoreTransposed::No >',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html',1,'dlaf::matrix']]], - ['panel_3c_20axis_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ayes_20_3e_317',['Panel< axis, const T, D, StoreTransposed::Yes >',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_yes_01_4.html',1,'dlaf::matrix']]], - ['panel_3c_20orthogonal_28axis_29_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ano_20_3e_318',['Panel< orthogonal(axis), const T, D, StoreTransposed::No >',['../structdlaf_1_1matrix_1_1_panel.html',1,'dlaf::matrix']]], - ['parentdistribution_319',['parentDistribution',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a66c9a42baab1741e5b562ddac6c716b2',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['partialtransform_320',['PartialTransform',['../classdlaf_1_1internal_1_1_partial_transform.html',1,'dlaf::internal']]], - ['partialtransformbase_321',['PartialTransformBase',['../structdlaf_1_1internal_1_1_partial_transform_base.html',1,'dlaf::internal']]], - ['partialtransformdetach_322',['PartialTransformDetach',['../classdlaf_1_1internal_1_1_partial_transform_detach.html',1,'dlaf::internal']]], - ['partialtransformmpi_323',['PartialTransformMPI',['../classdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i.html',1,'dlaf::comm::internal']]], - ['partialtransformmpibase_324',['PartialTransformMPIBase',['../structdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i_base.html',1,'dlaf::comm::internal']]], - ['partialtransformmpidetach_325',['PartialTransformMPIDetach',['../classdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i_detach.html',1,'dlaf::comm::internal']]], - ['permutations_326',['Permutations',['../structdlaf_1_1permutations_1_1internal_1_1_permutations.html',1,'dlaf::permutations::internal']]], - ['permute_327',['permute',['../permutations_2general_8h.html#af301d1e323dbafdc8503756ccd42edcd',1,'dlaf::permutations::permute(SizeType i_begin, SizeType i_end, Matrix< const SizeType, D > &perms, Matrix< const T, D > &mat_in, Matrix< T, D > &mat_out)'],['../permutations_2general_8h.html#a9dbcd748dd76ddf1b2dfef0d08bba03c',1,'dlaf::permutations::permute(comm::CommunicatorGrid grid, common::Pipeline< comm::Communicator > &sub_task_chain, SizeType i_begin, SizeType i_end, Matrix< const SizeType, D > &perms, Matrix< const T, D > &mat_in, Matrix< T, D > &mat_out)']]], - ['pipeline_328',['Pipeline',['../classdlaf_1_1common_1_1_pipeline.html',1,'dlaf::common::Pipeline< T >'],['../classdlaf_1_1common_1_1_pipeline.html#a6379e14f25ed8d98d3b3434ec0d54a62',1,'dlaf::common::Pipeline::Pipeline()']]], - ['pipeline_2eh_329',['pipeline.h',['../pipeline_8h.html',1,'']]], - ['policy_330',['Policy',['../classdlaf_1_1internal_1_1_policy.html',1,'dlaf::internal']]], - ['potrf_331',['potrf',['../lapack_2tile_8h.html#aab0795227301a9e8e63232028e331fe0',1,'dlaf::tile']]], - ['potrfinfo_332',['potrfInfo',['../lapack_2tile_8h.html#aa917a1efbd53213821776d7d632ee41c',1,'dlaf::tile']]], - ['prependpack_333',['PrependPack',['../structdlaf_1_1internal_1_1_prepend_pack.html',1,'dlaf::internal']]], - ['prependpack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20t_20_3e_334',['PrependPack< Pack< Ts... >, T >',['../structdlaf_1_1internal_1_1_prepend_pack_3_01_pack_3_01_ts_8_8_8_01_4_00_01_t_01_4.html',1,'dlaf::internal']]], - ['print_335',['print',['../print__csv_8h.html#ace48c87b6c1d3dc59bf005fb943c0629',1,'dlaf::matrix::print(format::csv, const Tile< const T, Device::CPU > &tile, std::ostream &os=std::cout)'],['../print__numpy_8h.html#ab467c4a97112f7b1d2cffe3acbb41643',1,'dlaf::matrix::print(format::numpy, const Tile< const T, Device::CPU > &tile, std::ostream &os=std::cout)'],['../print__csv_8h.html#ae6a4db345c18954cc7d83022a1a848d6',1,'dlaf::matrix::print(format::csv, std::string sym, Matrix< const T, Device::CPU > &mat, std::ostream &os=std::cout)'],['../print__gpu_8h.html#aa0268964c774b355f5513ae93a23b258',1,'dlaf::matrix::print(Format format, const Tile< const T, Device::GPU > &tile, std::ostream &os, whip::stream_t stream)']]], - ['print_5fcsv_2eh_336',['print_csv.h',['../print__csv_8h.html',1,'']]], - ['print_5fgpu_2eh_337',['print_gpu.h',['../print__gpu_8h.html',1,'']]], - ['print_5fnumpy_2eh_338',['print_numpy.h',['../print__numpy_8h.html',1,'']]], - ['ptr_339',['ptr',['../classdlaf_1_1matrix_1_1_tile.html#a9d16161523f90ce0f432af6e3eacb9f2',1,'dlaf::matrix::Tile::ptr(const TileElementIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_tile.html#ad14d471389721941e830f73d7ae0716d',1,'dlaf::matrix::Tile::ptr() const noexcept'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a7b8f0d5579d0f38743f123046780e860',1,'dlaf::matrix::Tile< const T, D >::ptr(const TileElementIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a018a34c9be26937cbb0c6dbb727c69ce',1,'dlaf::matrix::Tile< const T, D >::ptr() const noexcept']]] + ['p2p_2eh_313',['p2p.h',['../p2p_8h.html',1,'']]], + ['p2p_5fallsum_2eh_314',['p2p_allsum.h',['../p2p__allsum_8h.html',1,'']]], + ['panel_315',['Panel',['../structdlaf_1_1matrix_1_1_panel.html',1,'dlaf::matrix::Panel< axis, T, D, Storage >'],['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ad63a4d555cea1fdf2f6d5745125fe3cb',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::Panel()']]], + ['panel_2eh_316',['panel.h',['../panel_8h.html',1,'']]], + ['panel_3c_20axis_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ano_20_3e_317',['Panel< axis, const T, D, StoreTransposed::No >',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html',1,'dlaf::matrix']]], + ['panel_3c_20axis_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ayes_20_3e_318',['Panel< axis, const T, D, StoreTransposed::Yes >',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_yes_01_4.html',1,'dlaf::matrix']]], + ['panel_3c_20orthogonal_28axis_29_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ano_20_3e_319',['Panel< orthogonal(axis), const T, D, StoreTransposed::No >',['../structdlaf_1_1matrix_1_1_panel.html',1,'dlaf::matrix']]], + ['parentdistribution_320',['parentDistribution',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a66c9a42baab1741e5b562ddac6c716b2',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['partialtransform_321',['PartialTransform',['../classdlaf_1_1internal_1_1_partial_transform.html',1,'dlaf::internal']]], + ['partialtransformbase_322',['PartialTransformBase',['../structdlaf_1_1internal_1_1_partial_transform_base.html',1,'dlaf::internal']]], + ['partialtransformdetach_323',['PartialTransformDetach',['../classdlaf_1_1internal_1_1_partial_transform_detach.html',1,'dlaf::internal']]], + ['partialtransformmpi_324',['PartialTransformMPI',['../classdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i.html',1,'dlaf::comm::internal']]], + ['partialtransformmpibase_325',['PartialTransformMPIBase',['../structdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i_base.html',1,'dlaf::comm::internal']]], + ['partialtransformmpidetach_326',['PartialTransformMPIDetach',['../classdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i_detach.html',1,'dlaf::comm::internal']]], + ['permutations_327',['Permutations',['../structdlaf_1_1permutations_1_1internal_1_1_permutations.html',1,'dlaf::permutations::internal']]], + ['permute_328',['permute',['../permutations_2general_8h.html#af301d1e323dbafdc8503756ccd42edcd',1,'dlaf::permutations::permute(SizeType i_begin, SizeType i_end, Matrix< const SizeType, D > &perms, Matrix< const T, D > &mat_in, Matrix< T, D > &mat_out)'],['../permutations_2general_8h.html#a9dbcd748dd76ddf1b2dfef0d08bba03c',1,'dlaf::permutations::permute(comm::CommunicatorGrid grid, common::Pipeline< comm::Communicator > &sub_task_chain, SizeType i_begin, SizeType i_end, Matrix< const SizeType, D > &perms, Matrix< const T, D > &mat_in, Matrix< T, D > &mat_out)']]], + ['pipeline_329',['Pipeline',['../classdlaf_1_1common_1_1_pipeline.html',1,'dlaf::common::Pipeline< T >'],['../classdlaf_1_1common_1_1_pipeline.html#a6379e14f25ed8d98d3b3434ec0d54a62',1,'dlaf::common::Pipeline::Pipeline()']]], + ['pipeline_2eh_330',['pipeline.h',['../pipeline_8h.html',1,'']]], + ['policy_331',['Policy',['../classdlaf_1_1internal_1_1_policy.html',1,'dlaf::internal']]], + ['potrf_332',['potrf',['../lapack_2tile_8h.html#aab0795227301a9e8e63232028e331fe0',1,'dlaf::tile']]], + ['potrfinfo_333',['potrfInfo',['../lapack_2tile_8h.html#aa917a1efbd53213821776d7d632ee41c',1,'dlaf::tile']]], + ['prependpack_334',['PrependPack',['../structdlaf_1_1internal_1_1_prepend_pack.html',1,'dlaf::internal']]], + ['prependpack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20t_20_3e_335',['PrependPack< Pack< Ts... >, T >',['../structdlaf_1_1internal_1_1_prepend_pack_3_01_pack_3_01_ts_8_8_8_01_4_00_01_t_01_4.html',1,'dlaf::internal']]], + ['print_336',['print',['../print__csv_8h.html#ace48c87b6c1d3dc59bf005fb943c0629',1,'dlaf::matrix::print(format::csv, const Tile< const T, Device::CPU > &tile, std::ostream &os=std::cout)'],['../print__numpy_8h.html#ab467c4a97112f7b1d2cffe3acbb41643',1,'dlaf::matrix::print(format::numpy, const Tile< const T, Device::CPU > &tile, std::ostream &os=std::cout)'],['../print__csv_8h.html#ae6a4db345c18954cc7d83022a1a848d6',1,'dlaf::matrix::print(format::csv, std::string sym, Matrix< const T, Device::CPU > &mat, std::ostream &os=std::cout)'],['../print__gpu_8h.html#aa0268964c774b355f5513ae93a23b258',1,'dlaf::matrix::print(Format format, const Tile< const T, Device::GPU > &tile, std::ostream &os, whip::stream_t stream)']]], + ['print_5fcsv_2eh_337',['print_csv.h',['../print__csv_8h.html',1,'']]], + ['print_5fgpu_2eh_338',['print_gpu.h',['../print__gpu_8h.html',1,'']]], + ['print_5fnumpy_2eh_339',['print_numpy.h',['../print__numpy_8h.html',1,'']]], + ['ptr_340',['ptr',['../classdlaf_1_1matrix_1_1_tile.html#a9d16161523f90ce0f432af6e3eacb9f2',1,'dlaf::matrix::Tile::ptr(const TileElementIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_tile.html#ad14d471389721941e830f73d7ae0716d',1,'dlaf::matrix::Tile::ptr() const noexcept'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a7b8f0d5579d0f38743f123046780e860',1,'dlaf::matrix::Tile< const T, D >::ptr(const TileElementIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a018a34c9be26937cbb0c6dbb727c69ce',1,'dlaf::matrix::Tile< const T, D >::ptr() const noexcept']]] ]; diff --git a/master/search/all_f.js b/master/search/all_f.js index cd4991cf9a..530c767b13 100644 --- a/master/search/all_f.js +++ b/master/search/all_f.js @@ -1,6 +1,6 @@ var searchData= [ - ['qr_340',['QR',['../structdlaf_1_1factorization_1_1internal_1_1_q_r.html',1,'dlaf::factorization::internal']]], - ['qr_2eh_341',['qr.h',['../qr_8h.html',1,'']]], - ['qr_5ftfactor_342',['QR_Tfactor',['../structdlaf_1_1factorization_1_1internal_1_1_q_r___tfactor.html',1,'dlaf::factorization::internal']]] + ['qr_341',['QR',['../structdlaf_1_1factorization_1_1internal_1_1_q_r.html',1,'dlaf::factorization::internal']]], + ['qr_2eh_342',['qr.h',['../qr_8h.html',1,'']]], + ['qr_5ftfactor_343',['QR_Tfactor',['../structdlaf_1_1factorization_1_1internal_1_1_q_r___tfactor.html',1,'dlaf::factorization::internal']]] ]; diff --git a/master/search/classes_0.js b/master/search/classes_0.js index 8e0f647b5a..7bddff5f72 100644 --- a/master/search/classes_0.js +++ b/master/search/classes_0.js @@ -1,13 +1,13 @@ var searchData= [ - ['alwaysfalse_532',['AlwaysFalse',['../structdlaf_1_1internal_1_1_always_false.html',1,'dlaf::internal']]], - ['anyof_533',['AnyOf',['../structdlaf_1_1internal_1_1_any_of.html',1,'dlaf::internal']]], - ['anyof_3c_20std_3a_3ais_5fsame_3c_20t_2c_20ts_20_3e_2e_2e_2e_20_3e_534',['AnyOf< std::is_same< T, Ts >... >',['../structdlaf_1_1internal_1_1_any_of.html',1,'dlaf::internal']]], - ['anyof_3c_3e_535',['AnyOf<>',['../structdlaf_1_1internal_1_1_any_of_3_4.html',1,'dlaf::internal']]], - ['applyhhtodoubletilerow_536',['ApplyHHToDoubleTileRow',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_double_tile_row.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['applyhhtodoubletilerow_3c_20backend_3a_3agpu_2c_20t_20_3e_537',['ApplyHHToDoubleTileRow< Backend::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_double_tile_row_3_01_backend_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['applyhhtodoubletilerow_3c_20backend_3a_3amc_2c_20t_20_3e_538',['ApplyHHToDoubleTileRow< Backend::MC, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_double_tile_row_3_01_backend_1_1_m_c_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['applyhhtosingletilerow_539',['ApplyHHToSingleTileRow',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_single_tile_row.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['applyhhtosingletilerow_3c_20backend_3a_3agpu_2c_20t_20_3e_540',['ApplyHHToSingleTileRow< Backend::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_single_tile_row_3_01_backend_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['applyhhtosingletilerow_3c_20backend_3a_3amc_2c_20t_20_3e_541',['ApplyHHToSingleTileRow< Backend::MC, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_single_tile_row_3_01_backend_1_1_m_c_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]] + ['alwaysfalse_536',['AlwaysFalse',['../structdlaf_1_1internal_1_1_always_false.html',1,'dlaf::internal']]], + ['anyof_537',['AnyOf',['../structdlaf_1_1internal_1_1_any_of.html',1,'dlaf::internal']]], + ['anyof_3c_20std_3a_3ais_5fsame_3c_20t_2c_20ts_20_3e_2e_2e_2e_20_3e_538',['AnyOf< std::is_same< T, Ts >... >',['../structdlaf_1_1internal_1_1_any_of.html',1,'dlaf::internal']]], + ['anyof_3c_3e_539',['AnyOf<>',['../structdlaf_1_1internal_1_1_any_of_3_4.html',1,'dlaf::internal']]], + ['applyhhtodoubletilerow_540',['ApplyHHToDoubleTileRow',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_double_tile_row.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['applyhhtodoubletilerow_3c_20backend_3a_3agpu_2c_20t_20_3e_541',['ApplyHHToDoubleTileRow< Backend::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_double_tile_row_3_01_backend_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['applyhhtodoubletilerow_3c_20backend_3a_3amc_2c_20t_20_3e_542',['ApplyHHToDoubleTileRow< Backend::MC, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_double_tile_row_3_01_backend_1_1_m_c_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['applyhhtosingletilerow_543',['ApplyHHToSingleTileRow',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_single_tile_row.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['applyhhtosingletilerow_3c_20backend_3a_3agpu_2c_20t_20_3e_544',['ApplyHHToSingleTileRow< Backend::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_single_tile_row_3_01_backend_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['applyhhtosingletilerow_3c_20backend_3a_3amc_2c_20t_20_3e_545',['ApplyHHToSingleTileRow< Backend::MC, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_apply_h_h_to_single_tile_row_3_01_backend_1_1_m_c_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]] ]; diff --git a/master/search/classes_1.js b/master/search/classes_1.js index 03e94f2746..b03fff3d12 100644 --- a/master/search/classes_1.js +++ b/master/search/classes_1.js @@ -1,10 +1,10 @@ var searchData= [ - ['backtransformationreductiontoband_542',['BackTransformationReductionToBand',['../structdlaf_1_1eigensolver_1_1internal_1_1_back_transformation_reduction_to_band.html',1,'dlaf::eigensolver::internal']]], - ['backtransformationt2b_543',['BackTransformationT2B',['../structdlaf_1_1eigensolver_1_1internal_1_1_back_transformation_t2_b.html',1,'dlaf::eigensolver::internal']]], - ['bandblock_544',['BandBlock',['../classdlaf_1_1eigensolver_1_1internal_1_1_band_block.html',1,'dlaf::eigensolver::internal']]], - ['bandtotridiag_545',['BandToTridiag',['../structdlaf_1_1eigensolver_1_1internal_1_1_band_to_tridiag.html',1,'dlaf::eigensolver::internal']]], - ['bandtotridiag_3c_20backend_3a_3amc_2c_20d_2c_20t_20_3e_546',['BandToTridiag< Backend::MC, D, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1_band_to_tridiag_3_01_backend_1_1_m_c_00_01_d_00_01_t_01_4.html',1,'dlaf::eigensolver::internal']]], - ['basic_5fcoords_547',['basic_coords',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html',1,'dlaf::common::internal']]], - ['buffer_548',['Buffer',['../structdlaf_1_1common_1_1_buffer.html',1,'dlaf::common']]] + ['backtransformationreductiontoband_546',['BackTransformationReductionToBand',['../structdlaf_1_1eigensolver_1_1internal_1_1_back_transformation_reduction_to_band.html',1,'dlaf::eigensolver::internal']]], + ['backtransformationt2b_547',['BackTransformationT2B',['../structdlaf_1_1eigensolver_1_1internal_1_1_back_transformation_t2_b.html',1,'dlaf::eigensolver::internal']]], + ['bandblock_548',['BandBlock',['../classdlaf_1_1eigensolver_1_1internal_1_1_band_block.html',1,'dlaf::eigensolver::internal']]], + ['bandtotridiag_549',['BandToTridiag',['../structdlaf_1_1eigensolver_1_1internal_1_1_band_to_tridiag.html',1,'dlaf::eigensolver::internal']]], + ['bandtotridiag_3c_20backend_3a_3amc_2c_20d_2c_20t_20_3e_550',['BandToTridiag< Backend::MC, D, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1_band_to_tridiag_3_01_backend_1_1_m_c_00_01_d_00_01_t_01_4.html',1,'dlaf::eigensolver::internal']]], + ['basic_5fcoords_551',['basic_coords',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html',1,'dlaf::common::internal']]], + ['buffer_552',['Buffer',['../structdlaf_1_1common_1_1_buffer.html',1,'dlaf::common']]] ]; diff --git a/master/search/classes_10.js b/master/search/classes_10.js index 692cecfae6..4f6c9d02fd 100644 --- a/master/search/classes_10.js +++ b/master/search/classes_10.js @@ -1,23 +1,23 @@ var searchData= [ - ['tile_702',['Tile',['../classdlaf_1_1matrix_1_1_tile.html',1,'dlaf::matrix']]], - ['tile_3c_20const_20t_2c_20d_20_3e_703',['Tile< const T, D >',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html',1,'dlaf::matrix']]], - ['tileaccesshelper_704',['TileAccessHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_tile_access_helper.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['tilecollector_705',['TileCollector',['../classdlaf_1_1eigensolver_1_1internal_1_1_tile_collector.html',1,'dlaf::eigensolver::internal']]], - ['tiledata_706',['TileData',['../classdlaf_1_1matrix_1_1internal_1_1_tile_data.html',1,'dlaf::matrix::internal']]], - ['tilepipeline_707',['TilePipeline',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html',1,'dlaf::matrix::internal']]], - ['timer_708',['Timer',['../classdlaf_1_1common_1_1_timer.html',1,'dlaf::common']]], - ['transformpack_709',['TransformPack',['../structdlaf_1_1internal_1_1_transform_pack.html',1,'dlaf::internal']]], - ['transformpack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20transformer_20_3e_710',['TransformPack< Pack< Ts... >, Transformer >',['../structdlaf_1_1internal_1_1_transform_pack_3_01_pack_3_01_ts_8_8_8_01_4_00_01_transformer_01_4.html',1,'dlaf::internal']]], - ['triangular_711',['Triangular',['../structdlaf_1_1multiplication_1_1internal_1_1_triangular.html',1,'dlaf::multiplication::internal::Triangular< backend, device, T >'],['../structdlaf_1_1solver_1_1internal_1_1_triangular.html',1,'dlaf::solver::internal::Triangular< backend, device, T >']]], - ['tridiagresult_712',['TridiagResult',['../structdlaf_1_1eigensolver_1_1_tridiag_result.html',1,'dlaf::eigensolver']]], - ['tridiagsolver_713',['TridiagSolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_tridiag_solver.html',1,'dlaf::eigensolver::internal']]], - ['trmmsizes_714',['trmmSizes',['../structdlaf_1_1tile_1_1internal_1_1trmm_sizes.html',1,'dlaf::tile::internal']]], - ['trsmsizes_715',['trsmSizes',['../structdlaf_1_1tile_1_1internal_1_1trsm_sizes.html',1,'dlaf::tile::internal']]], - ['tuneparameters_716',['TuneParameters',['../structdlaf_1_1_tune_parameters.html',1,'dlaf']]], - ['type_5fhandler_717',['type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html',1,'dlaf::comm::internal']]], - ['typeinfo_718',['TypeInfo',['../structdlaf_1_1_type_info.html',1,'dlaf']]], - ['typeinfo_3c_20const_20t_20_3e_719',['TypeInfo< const T >',['../structdlaf_1_1_type_info_3_01const_01_t_01_4.html',1,'dlaf']]], - ['typeinfo_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_720',['TypeInfo< std::complex< T > >',['../structdlaf_1_1_type_info_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf']]], - ['typelist_721',['TypeList',['../structdlaf_1_1internal_1_1_type_list.html',1,'dlaf::internal']]] + ['tile_705',['Tile',['../classdlaf_1_1matrix_1_1_tile.html',1,'dlaf::matrix']]], + ['tile_3c_20const_20t_2c_20d_20_3e_706',['Tile< const T, D >',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html',1,'dlaf::matrix']]], + ['tileaccesshelper_707',['TileAccessHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_tile_access_helper.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['tilecollector_708',['TileCollector',['../classdlaf_1_1eigensolver_1_1internal_1_1_tile_collector.html',1,'dlaf::eigensolver::internal']]], + ['tiledata_709',['TileData',['../classdlaf_1_1matrix_1_1internal_1_1_tile_data.html',1,'dlaf::matrix::internal']]], + ['tilepipeline_710',['TilePipeline',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html',1,'dlaf::matrix::internal']]], + ['timer_711',['Timer',['../classdlaf_1_1common_1_1_timer.html',1,'dlaf::common']]], + ['transformpack_712',['TransformPack',['../structdlaf_1_1internal_1_1_transform_pack.html',1,'dlaf::internal']]], + ['transformpack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20transformer_20_3e_713',['TransformPack< Pack< Ts... >, Transformer >',['../structdlaf_1_1internal_1_1_transform_pack_3_01_pack_3_01_ts_8_8_8_01_4_00_01_transformer_01_4.html',1,'dlaf::internal']]], + ['triangular_714',['Triangular',['../structdlaf_1_1multiplication_1_1internal_1_1_triangular.html',1,'dlaf::multiplication::internal::Triangular< backend, device, T >'],['../structdlaf_1_1solver_1_1internal_1_1_triangular.html',1,'dlaf::solver::internal::Triangular< backend, device, T >']]], + ['tridiagresult_715',['TridiagResult',['../structdlaf_1_1eigensolver_1_1_tridiag_result.html',1,'dlaf::eigensolver']]], + ['tridiagsolver_716',['TridiagSolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_tridiag_solver.html',1,'dlaf::eigensolver::internal']]], + ['trmmsizes_717',['trmmSizes',['../structdlaf_1_1tile_1_1internal_1_1trmm_sizes.html',1,'dlaf::tile::internal']]], + ['trsmsizes_718',['trsmSizes',['../structdlaf_1_1tile_1_1internal_1_1trsm_sizes.html',1,'dlaf::tile::internal']]], + ['tuneparameters_719',['TuneParameters',['../structdlaf_1_1_tune_parameters.html',1,'dlaf']]], + ['type_5fhandler_720',['type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html',1,'dlaf::comm::internal']]], + ['typeinfo_721',['TypeInfo',['../structdlaf_1_1_type_info.html',1,'dlaf']]], + ['typeinfo_3c_20const_20t_20_3e_722',['TypeInfo< const T >',['../structdlaf_1_1_type_info_3_01const_01_t_01_4.html',1,'dlaf']]], + ['typeinfo_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_723',['TypeInfo< std::complex< T > >',['../structdlaf_1_1_type_info_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf']]], + ['typelist_724',['TypeList',['../structdlaf_1_1internal_1_1_type_list.html',1,'dlaf::internal']]] ]; diff --git a/master/search/classes_11.js b/master/search/classes_11.js index 48045b40a3..c9242da99e 100644 --- a/master/search/classes_11.js +++ b/master/search/classes_11.js @@ -1,19 +1,19 @@ var searchData= [ - ['unique_722',['Unique',['../structdlaf_1_1internal_1_1_unique.html',1,'dlaf::internal']]], - ['unique_3c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_723',['Unique< Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_3_01_pack_3_01_ts_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], - ['uniquehelper_724',['UniqueHelper',['../structdlaf_1_1internal_1_1_unique_helper.html',1,'dlaf::internal']]], - ['uniquehelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_20u_2c_20us_2e_2e_2e_20_3e_20_3e_725',['UniqueHelper< Pack< Ts... >, Pack< U, Us... > >',['../structdlaf_1_1internal_1_1_unique_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_01_u_00_01_us_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], - ['uniquehelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_3e_20_3e_726',['UniqueHelper< Pack< Ts... >, Pack<> >',['../structdlaf_1_1internal_1_1_unique_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_4_01_4.html',1,'dlaf::internal']]], - ['uniquehelper_3c_20pack_3c_3e_2c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_727',['UniqueHelper< Pack<>, Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_helper.html',1,'dlaf::internal']]], - ['uniquepack_728',['UniquePack',['../structdlaf_1_1internal_1_1_unique_pack.html',1,'dlaf::internal']]], - ['uniquepack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_729',['UniquePack< Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_pack_3_01_pack_3_01_ts_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], - ['uniquepackhelper_730',['UniquePackHelper',['../structdlaf_1_1internal_1_1_unique_pack_helper.html',1,'dlaf::internal']]], - ['uniquepackhelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_20u_2c_20us_2e_2e_2e_20_3e_20_3e_731',['UniquePackHelper< Pack< Ts... >, Pack< U, Us... > >',['../structdlaf_1_1internal_1_1_unique_pack_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_01_u_00_01_us_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], - ['uniquepackhelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_3e_20_3e_732',['UniquePackHelper< Pack< Ts... >, Pack<> >',['../structdlaf_1_1internal_1_1_unique_pack_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_4_01_4.html',1,'dlaf::internal']]], - ['uniquepackhelper_3c_20pack_3c_3e_2c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_733',['UniquePackHelper< Pack<>, Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_pack_helper.html',1,'dlaf::internal']]], - ['unwrapper_734',['Unwrapper',['../structdlaf_1_1common_1_1internal_1_1_unwrapper.html',1,'dlaf::common::internal']]], - ['unwrapper_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20t1_2c_20t2_2c_20at_20_3e_20_3e_735',['Unwrapper< pika::execution::experimental::async_rw_mutex_access_wrapper< T1, T2, at > >',['../structdlaf_1_1common_1_1internal_1_1_unwrapper_3_01pika_1_1execution_1_1experimental_1_1async__r5be7b0afcb71e6df73ebcc954d80821d.html',1,'dlaf::common::internal']]], - ['unwrapper_3c_20std_3a_3areference_5fwrapper_3c_20t_20_3e_20_3e_736',['Unwrapper< std::reference_wrapper< T > >',['../structdlaf_1_1common_1_1internal_1_1_unwrapper_3_01std_1_1reference__wrapper_3_01_t_01_4_01_4.html',1,'dlaf::common::internal']]], - ['unwrapping_737',['Unwrapping',['../structdlaf_1_1common_1_1internal_1_1_unwrapping.html',1,'dlaf::common::internal']]] + ['unique_725',['Unique',['../structdlaf_1_1internal_1_1_unique.html',1,'dlaf::internal']]], + ['unique_3c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_726',['Unique< Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_3_01_pack_3_01_ts_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], + ['uniquehelper_727',['UniqueHelper',['../structdlaf_1_1internal_1_1_unique_helper.html',1,'dlaf::internal']]], + ['uniquehelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_20u_2c_20us_2e_2e_2e_20_3e_20_3e_728',['UniqueHelper< Pack< Ts... >, Pack< U, Us... > >',['../structdlaf_1_1internal_1_1_unique_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_01_u_00_01_us_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], + ['uniquehelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_3e_20_3e_729',['UniqueHelper< Pack< Ts... >, Pack<> >',['../structdlaf_1_1internal_1_1_unique_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_4_01_4.html',1,'dlaf::internal']]], + ['uniquehelper_3c_20pack_3c_3e_2c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_730',['UniqueHelper< Pack<>, Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_helper.html',1,'dlaf::internal']]], + ['uniquepack_731',['UniquePack',['../structdlaf_1_1internal_1_1_unique_pack.html',1,'dlaf::internal']]], + ['uniquepack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_732',['UniquePack< Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_pack_3_01_pack_3_01_ts_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], + ['uniquepackhelper_733',['UniquePackHelper',['../structdlaf_1_1internal_1_1_unique_pack_helper.html',1,'dlaf::internal']]], + ['uniquepackhelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_20u_2c_20us_2e_2e_2e_20_3e_20_3e_734',['UniquePackHelper< Pack< Ts... >, Pack< U, Us... > >',['../structdlaf_1_1internal_1_1_unique_pack_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_01_u_00_01_us_8_8_8_01_4_01_4.html',1,'dlaf::internal']]], + ['uniquepackhelper_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20pack_3c_3e_20_3e_735',['UniquePackHelper< Pack< Ts... >, Pack<> >',['../structdlaf_1_1internal_1_1_unique_pack_helper_3_01_pack_3_01_ts_8_8_8_01_4_00_01_pack_3_4_01_4.html',1,'dlaf::internal']]], + ['uniquepackhelper_3c_20pack_3c_3e_2c_20pack_3c_20ts_2e_2e_2e_20_3e_20_3e_736',['UniquePackHelper< Pack<>, Pack< Ts... > >',['../structdlaf_1_1internal_1_1_unique_pack_helper.html',1,'dlaf::internal']]], + ['unwrapper_737',['Unwrapper',['../structdlaf_1_1common_1_1internal_1_1_unwrapper.html',1,'dlaf::common::internal']]], + ['unwrapper_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20t1_2c_20t2_2c_20at_20_3e_20_3e_738',['Unwrapper< pika::execution::experimental::async_rw_mutex_access_wrapper< T1, T2, at > >',['../structdlaf_1_1common_1_1internal_1_1_unwrapper_3_01pika_1_1execution_1_1experimental_1_1async__r5be7b0afcb71e6df73ebcc954d80821d.html',1,'dlaf::common::internal']]], + ['unwrapper_3c_20std_3a_3areference_5fwrapper_3c_20t_20_3e_20_3e_739',['Unwrapper< std::reference_wrapper< T > >',['../structdlaf_1_1common_1_1internal_1_1_unwrapper_3_01std_1_1reference__wrapper_3_01_t_01_4_01_4.html',1,'dlaf::common::internal']]], + ['unwrapping_740',['Unwrapping',['../structdlaf_1_1common_1_1internal_1_1_unwrapping.html',1,'dlaf::common::internal']]] ]; diff --git a/master/search/classes_12.js b/master/search/classes_12.js index 1087e89a4c..ed3bcf5e6b 100644 --- a/master/search/classes_12.js +++ b/master/search/classes_12.js @@ -1,7 +1,7 @@ var searchData= [ - ['vaccesshelper_738',['VAccessHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1_v_access_helper.html',1,'dlaf::eigensolver::internal']]], - ['vector_739',['vector',['../structdlaf_1_1common_1_1internal_1_1vector.html',1,'dlaf::common::internal']]], - ['vector_3c_20readonlysendertype_20_3e_740',['vector< ReadOnlySenderType >',['../structdlaf_1_1common_1_1internal_1_1vector.html',1,'dlaf::common::internal']]], - ['view_741',['View',['../structdlaf_1_1matrix_1_1internal_1_1_view.html',1,'dlaf::matrix::internal']]] + ['vaccesshelper_741',['VAccessHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1_v_access_helper.html',1,'dlaf::eigensolver::internal']]], + ['vector_742',['vector',['../structdlaf_1_1common_1_1internal_1_1vector.html',1,'dlaf::common::internal']]], + ['vector_3c_20readonlysendertype_20_3e_743',['vector< ReadOnlySenderType >',['../structdlaf_1_1common_1_1internal_1_1vector.html',1,'dlaf::common::internal']]], + ['view_744',['View',['../structdlaf_1_1matrix_1_1internal_1_1_view.html',1,'dlaf::matrix::internal']]] ]; diff --git a/master/search/classes_13.js b/master/search/classes_13.js index b6be8a7824..0b08f1ea1f 100644 --- a/master/search/classes_13.js +++ b/master/search/classes_13.js @@ -1,6 +1,6 @@ var searchData= [ - ['workspace_742',['WorkSpace',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space.html',1,'dlaf::eigensolver::internal']]], - ['workspacehost_743',['WorkSpaceHost',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space_host.html',1,'dlaf::eigensolver::internal']]], - ['workspacehostmirror_744',['WorkSpaceHostMirror',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space_host_mirror.html',1,'dlaf::eigensolver::internal']]] + ['workspace_745',['WorkSpace',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space.html',1,'dlaf::eigensolver::internal']]], + ['workspacehost_746',['WorkSpaceHost',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space_host.html',1,'dlaf::eigensolver::internal']]], + ['workspacehostmirror_747',['WorkSpaceHostMirror',['../structdlaf_1_1eigensolver_1_1internal_1_1_work_space_host_mirror.html',1,'dlaf::eigensolver::internal']]] ]; diff --git a/master/search/classes_2.js b/master/search/classes_2.js index e4bd5fede9..0f6c1a6b5f 100644 --- a/master/search/classes_2.js +++ b/master/search/classes_2.js @@ -1,33 +1,33 @@ var searchData= [ - ['cholesky_549',['Cholesky',['../structdlaf_1_1factorization_1_1internal_1_1_cholesky.html',1,'dlaf::factorization::internal']]], - ['communicationdevice_550',['CommunicationDevice',['../structdlaf_1_1comm_1_1_communication_device.html',1,'dlaf::comm']]], - ['communicationdevice_3c_20device_3a_3agpu_20_3e_551',['CommunicationDevice< Device::GPU >',['../structdlaf_1_1comm_1_1_communication_device_3_01_device_1_1_g_p_u_01_4.html',1,'dlaf::comm']]], - ['communicator_552',['Communicator',['../classdlaf_1_1comm_1_1_communicator.html',1,'dlaf::comm']]], - ['communicatorgrid_553',['CommunicatorGrid',['../classdlaf_1_1comm_1_1_communicator_grid.html',1,'dlaf::comm']]], - ['computepanelhelper_554',['ComputePanelHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper.html',1,'dlaf::eigensolver::internal::red2band']]], - ['computepanelhelper_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_555',['ComputePanelHelper< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_e566a066992e6980ee13cab3c8831bc5.html',1,'dlaf::eigensolver::internal::red2band']]], - ['computepanelhelper_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_556',['ComputePanelHelper< Backend::MC, Device::CPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_f6049763c467d19a4a2bae31f1d19649.html',1,'dlaf::eigensolver::internal::red2band']]], - ['configuration_557',['configuration',['../structdlaf_1_1configuration.html',1,'dlaf']]], - ['contains_558',['Contains',['../structdlaf_1_1internal_1_1_contains.html',1,'dlaf::internal']]], - ['contains_3c_20t_20_3e_559',['Contains< T >',['../structdlaf_1_1internal_1_1_contains_3_01_t_01_4.html',1,'dlaf::internal']]], - ['contains_3c_20t_2c_20t_2c_20ts_2e_2e_2e_20_3e_560',['Contains< T, T, Ts... >',['../structdlaf_1_1internal_1_1_contains_3_01_t_00_01_t_00_01_ts_8_8_8_01_4.html',1,'dlaf::internal']]], - ['contains_3c_20t_2c_20ts_2e_2e_2e_20_3e_561',['Contains< T, Ts... >',['../structdlaf_1_1internal_1_1_contains.html',1,'dlaf::internal']]], - ['contains_3c_20t_2c_20u_2c_20ts_2e_2e_2e_20_3e_562',['Contains< T, U, Ts... >',['../structdlaf_1_1internal_1_1_contains_3_01_t_00_01_u_00_01_ts_8_8_8_01_4.html',1,'dlaf::internal']]], - ['copybackend_563',['CopyBackend',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend.html',1,'dlaf::matrix::internal']]], - ['copybackend_3c_20device_3a_3acpu_2c_20device_3a_3acpu_20_3e_564',['CopyBackend< Device::CPU, Device::CPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend_3_01_device_1_1_c_p_u_00_01_device_1_1_c_p_u_01_4.html',1,'dlaf::matrix::internal']]], - ['copybackend_3c_20device_3a_3acpu_2c_20device_3a_3agpu_20_3e_565',['CopyBackend< Device::CPU, Device::GPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend_3_01_device_1_1_c_p_u_00_01_device_1_1_g_p_u_01_4.html',1,'dlaf::matrix::internal']]], - ['copybackend_3c_20device_3a_3agpu_2c_20device_3a_3acpu_20_3e_566',['CopyBackend< Device::GPU, Device::CPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend_3_01_device_1_1_g_p_u_00_01_device_1_1_c_p_u_01_4.html',1,'dlaf::matrix::internal']]], - ['copybackend_3c_20device_3a_3agpu_2c_20device_3a_3agpu_20_3e_567',['CopyBackend< Device::GPU, Device::GPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend_3_01_device_1_1_g_p_u_00_01_device_1_1_g_p_u_01_4.html',1,'dlaf::matrix::internal']]], - ['copytile_568',['CopyTile',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile.html',1,'dlaf::matrix::internal']]], - ['copytile_3c_20t_2c_20device_3a_3acpu_2c_20device_3a_3acpu_20_3e_569',['CopyTile< T, Device::CPU, Device::CPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile_3_01_t_00_01_device_1_1_c_p_u_00_01_device_1_1_c_p_u_01_4.html',1,'dlaf::matrix::internal']]], - ['copytile_3c_20t_2c_20device_3a_3acpu_2c_20device_3a_3agpu_20_3e_570',['CopyTile< T, Device::CPU, Device::GPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile_3_01_t_00_01_device_1_1_c_p_u_00_01_device_1_1_g_p_u_01_4.html',1,'dlaf::matrix::internal']]], - ['copytile_3c_20t_2c_20device_3a_3agpu_2c_20device_3a_3acpu_20_3e_571',['CopyTile< T, Device::GPU, Device::CPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile_3_01_t_00_01_device_1_1_g_p_u_00_01_device_1_1_c_p_u_01_4.html',1,'dlaf::matrix::internal']]], - ['copytile_3c_20t_2c_20device_3a_3agpu_2c_20device_3a_3agpu_20_3e_572',['CopyTile< T, Device::GPU, Device::GPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile_3_01_t_00_01_device_1_1_g_p_u_00_01_device_1_1_g_p_u_01_4.html',1,'dlaf::matrix::internal']]], - ['cpptocudatype_573',['CppToCudaType',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type.html',1,'dlaf::util::internal']]], - ['cpptocudatype_3c_20const_20t_20_3e_574',['CppToCudaType< const T >',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type_3_01const_01_t_01_4.html',1,'dlaf::util::internal']]], - ['cpptocudatype_3c_20std_3a_3acomplex_3c_20double_20_3e_20_3e_575',['CppToCudaType< std::complex< double > >',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type_3_01std_1_1complex_3_01double_01_4_01_4.html',1,'dlaf::util::internal']]], - ['cpptocudatype_3c_20std_3a_3acomplex_3c_20float_20_3e_20_3e_576',['CppToCudaType< std::complex< float > >',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type_3_01std_1_1complex_3_01float_01_4_01_4.html',1,'dlaf::util::internal']]], - ['cpptocudatype_3c_20t_20_2a_20_3e_577',['CppToCudaType< T * >',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type_3_01_t_01_5_01_4.html',1,'dlaf::util::internal']]], - ['csv_578',['csv',['../structdlaf_1_1format_1_1csv.html',1,'dlaf::format']]] + ['cholesky_553',['Cholesky',['../structdlaf_1_1factorization_1_1internal_1_1_cholesky.html',1,'dlaf::factorization::internal']]], + ['communicationdevice_554',['CommunicationDevice',['../structdlaf_1_1comm_1_1_communication_device.html',1,'dlaf::comm']]], + ['communicationdevice_3c_20device_3a_3agpu_20_3e_555',['CommunicationDevice< Device::GPU >',['../structdlaf_1_1comm_1_1_communication_device_3_01_device_1_1_g_p_u_01_4.html',1,'dlaf::comm']]], + ['communicator_556',['Communicator',['../classdlaf_1_1comm_1_1_communicator.html',1,'dlaf::comm']]], + ['communicatorgrid_557',['CommunicatorGrid',['../classdlaf_1_1comm_1_1_communicator_grid.html',1,'dlaf::comm']]], + ['computepanelhelper_558',['ComputePanelHelper',['../structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper.html',1,'dlaf::eigensolver::internal::red2band']]], + ['computepanelhelper_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_559',['ComputePanelHelper< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_e566a066992e6980ee13cab3c8831bc5.html',1,'dlaf::eigensolver::internal::red2band']]], + ['computepanelhelper_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_560',['ComputePanelHelper< Backend::MC, Device::CPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_f6049763c467d19a4a2bae31f1d19649.html',1,'dlaf::eigensolver::internal::red2band']]], + ['configuration_561',['configuration',['../structdlaf_1_1configuration.html',1,'dlaf']]], + ['contains_562',['Contains',['../structdlaf_1_1internal_1_1_contains.html',1,'dlaf::internal']]], + ['contains_3c_20t_20_3e_563',['Contains< T >',['../structdlaf_1_1internal_1_1_contains_3_01_t_01_4.html',1,'dlaf::internal']]], + ['contains_3c_20t_2c_20t_2c_20ts_2e_2e_2e_20_3e_564',['Contains< T, T, Ts... >',['../structdlaf_1_1internal_1_1_contains_3_01_t_00_01_t_00_01_ts_8_8_8_01_4.html',1,'dlaf::internal']]], + ['contains_3c_20t_2c_20ts_2e_2e_2e_20_3e_565',['Contains< T, Ts... >',['../structdlaf_1_1internal_1_1_contains.html',1,'dlaf::internal']]], + ['contains_3c_20t_2c_20u_2c_20ts_2e_2e_2e_20_3e_566',['Contains< T, U, Ts... >',['../structdlaf_1_1internal_1_1_contains_3_01_t_00_01_u_00_01_ts_8_8_8_01_4.html',1,'dlaf::internal']]], + ['copybackend_567',['CopyBackend',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend.html',1,'dlaf::matrix::internal']]], + ['copybackend_3c_20device_3a_3acpu_2c_20device_3a_3acpu_20_3e_568',['CopyBackend< Device::CPU, Device::CPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend_3_01_device_1_1_c_p_u_00_01_device_1_1_c_p_u_01_4.html',1,'dlaf::matrix::internal']]], + ['copybackend_3c_20device_3a_3acpu_2c_20device_3a_3agpu_20_3e_569',['CopyBackend< Device::CPU, Device::GPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend_3_01_device_1_1_c_p_u_00_01_device_1_1_g_p_u_01_4.html',1,'dlaf::matrix::internal']]], + ['copybackend_3c_20device_3a_3agpu_2c_20device_3a_3acpu_20_3e_570',['CopyBackend< Device::GPU, Device::CPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend_3_01_device_1_1_g_p_u_00_01_device_1_1_c_p_u_01_4.html',1,'dlaf::matrix::internal']]], + ['copybackend_3c_20device_3a_3agpu_2c_20device_3a_3agpu_20_3e_571',['CopyBackend< Device::GPU, Device::GPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_backend_3_01_device_1_1_g_p_u_00_01_device_1_1_g_p_u_01_4.html',1,'dlaf::matrix::internal']]], + ['copytile_572',['CopyTile',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile.html',1,'dlaf::matrix::internal']]], + ['copytile_3c_20t_2c_20device_3a_3acpu_2c_20device_3a_3acpu_20_3e_573',['CopyTile< T, Device::CPU, Device::CPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile_3_01_t_00_01_device_1_1_c_p_u_00_01_device_1_1_c_p_u_01_4.html',1,'dlaf::matrix::internal']]], + ['copytile_3c_20t_2c_20device_3a_3acpu_2c_20device_3a_3agpu_20_3e_574',['CopyTile< T, Device::CPU, Device::GPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile_3_01_t_00_01_device_1_1_c_p_u_00_01_device_1_1_g_p_u_01_4.html',1,'dlaf::matrix::internal']]], + ['copytile_3c_20t_2c_20device_3a_3agpu_2c_20device_3a_3acpu_20_3e_575',['CopyTile< T, Device::GPU, Device::CPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile_3_01_t_00_01_device_1_1_g_p_u_00_01_device_1_1_c_p_u_01_4.html',1,'dlaf::matrix::internal']]], + ['copytile_3c_20t_2c_20device_3a_3agpu_2c_20device_3a_3agpu_20_3e_576',['CopyTile< T, Device::GPU, Device::GPU >',['../structdlaf_1_1matrix_1_1internal_1_1_copy_tile_3_01_t_00_01_device_1_1_g_p_u_00_01_device_1_1_g_p_u_01_4.html',1,'dlaf::matrix::internal']]], + ['cpptocudatype_577',['CppToCudaType',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type.html',1,'dlaf::util::internal']]], + ['cpptocudatype_3c_20const_20t_20_3e_578',['CppToCudaType< const T >',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type_3_01const_01_t_01_4.html',1,'dlaf::util::internal']]], + ['cpptocudatype_3c_20std_3a_3acomplex_3c_20double_20_3e_20_3e_579',['CppToCudaType< std::complex< double > >',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type_3_01std_1_1complex_3_01double_01_4_01_4.html',1,'dlaf::util::internal']]], + ['cpptocudatype_3c_20std_3a_3acomplex_3c_20float_20_3e_20_3e_580',['CppToCudaType< std::complex< float > >',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type_3_01std_1_1complex_3_01float_01_4_01_4.html',1,'dlaf::util::internal']]], + ['cpptocudatype_3c_20t_20_2a_20_3e_581',['CppToCudaType< T * >',['../structdlaf_1_1util_1_1internal_1_1_cpp_to_cuda_type_3_01_t_01_5_01_4.html',1,'dlaf::util::internal']]], + ['csv_582',['csv',['../structdlaf_1_1format_1_1csv.html',1,'dlaf::format']]] ]; diff --git a/master/search/classes_3.js b/master/search/classes_3.js index c6af2f3597..bc8429b743 100644 --- a/master/search/classes_3.js +++ b/master/search/classes_3.js @@ -1,19 +1,19 @@ var searchData= [ - ['data_5ftraits_579',['data_traits',['../structdlaf_1_1common_1_1data__traits.html',1,'dlaf::common']]], - ['data_5ftraits_3c_20buffer_3c_20t_20_3e_20_3e_580',['data_traits< Buffer< T > >',['../structdlaf_1_1common_1_1data__traits_3_01_buffer_3_01_t_01_4_01_4.html',1,'dlaf::common']]], - ['data_5ftraits_3c_20datadescriptor_3c_20t_20_3e_20_3e_581',['data_traits< DataDescriptor< T > >',['../structdlaf_1_1common_1_1data__traits_3_01_data_descriptor_3_01_t_01_4_01_4.html',1,'dlaf::common']]], - ['data_5ftraits_3c_20datadescriptor_3c_20t_5bn_5d_3e_20_3e_582',['data_traits< DataDescriptor< T[N]> >',['../structdlaf_1_1common_1_1data__traits_3_01_data_descriptor_3_01_t_0f_n_0e_4_01_4.html',1,'dlaf::common']]], - ['datadescriptor_583',['DataDescriptor',['../structdlaf_1_1common_1_1_data_descriptor.html',1,'dlaf::common']]], - ['datadescriptor_3c_20t_5bn_5d_3e_584',['DataDescriptor< T[N]>',['../structdlaf_1_1common_1_1_data_descriptor_3_01_t_0f_n_0e_4.html',1,'dlaf::common']]], - ['defaultbackend_585',['DefaultBackend',['../structdlaf_1_1_default_backend.html',1,'dlaf']]], - ['defaultbackend_3c_20device_3a_3acpu_20_3e_586',['DefaultBackend< Device::CPU >',['../structdlaf_1_1_default_backend_3_01_device_1_1_c_p_u_01_4.html',1,'dlaf']]], - ['defaultbackend_3c_20device_3a_3agpu_20_3e_587',['DefaultBackend< Device::GPU >',['../structdlaf_1_1_default_backend_3_01_device_1_1_g_p_u_01_4.html',1,'dlaf']]], - ['defaultdevice_588',['DefaultDevice',['../structdlaf_1_1_default_device.html',1,'dlaf']]], - ['defaultdevice_3c_20backend_3a_3agpu_20_3e_589',['DefaultDevice< Backend::GPU >',['../structdlaf_1_1_default_device_3_01_backend_1_1_g_p_u_01_4.html',1,'dlaf']]], - ['defaultdevice_3c_20backend_3a_3amc_20_3e_590',['DefaultDevice< Backend::MC >',['../structdlaf_1_1_default_device_3_01_backend_1_1_m_c_01_4.html',1,'dlaf']]], - ['distindexing_591',['DistIndexing',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_dist_indexing.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['distribution_592',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html',1,'dlaf::matrix']]], - ['distworkspacehostmirror_593',['DistWorkSpaceHostMirror',['../structdlaf_1_1eigensolver_1_1internal_1_1_dist_work_space_host_mirror.html',1,'dlaf::eigensolver::internal']]], - ['duplicatenocopy_594',['DuplicateNoCopy',['../structdlaf_1_1matrix_1_1_duplicate_no_copy.html',1,'dlaf::matrix']]] + ['data_5ftraits_583',['data_traits',['../structdlaf_1_1common_1_1data__traits.html',1,'dlaf::common']]], + ['data_5ftraits_3c_20buffer_3c_20t_20_3e_20_3e_584',['data_traits< Buffer< T > >',['../structdlaf_1_1common_1_1data__traits_3_01_buffer_3_01_t_01_4_01_4.html',1,'dlaf::common']]], + ['data_5ftraits_3c_20datadescriptor_3c_20t_20_3e_20_3e_585',['data_traits< DataDescriptor< T > >',['../structdlaf_1_1common_1_1data__traits_3_01_data_descriptor_3_01_t_01_4_01_4.html',1,'dlaf::common']]], + ['data_5ftraits_3c_20datadescriptor_3c_20t_5bn_5d_3e_20_3e_586',['data_traits< DataDescriptor< T[N]> >',['../structdlaf_1_1common_1_1data__traits_3_01_data_descriptor_3_01_t_0f_n_0e_4_01_4.html',1,'dlaf::common']]], + ['datadescriptor_587',['DataDescriptor',['../structdlaf_1_1common_1_1_data_descriptor.html',1,'dlaf::common']]], + ['datadescriptor_3c_20t_5bn_5d_3e_588',['DataDescriptor< T[N]>',['../structdlaf_1_1common_1_1_data_descriptor_3_01_t_0f_n_0e_4.html',1,'dlaf::common']]], + ['defaultbackend_589',['DefaultBackend',['../structdlaf_1_1_default_backend.html',1,'dlaf']]], + ['defaultbackend_3c_20device_3a_3acpu_20_3e_590',['DefaultBackend< Device::CPU >',['../structdlaf_1_1_default_backend_3_01_device_1_1_c_p_u_01_4.html',1,'dlaf']]], + ['defaultbackend_3c_20device_3a_3agpu_20_3e_591',['DefaultBackend< Device::GPU >',['../structdlaf_1_1_default_backend_3_01_device_1_1_g_p_u_01_4.html',1,'dlaf']]], + ['defaultdevice_592',['DefaultDevice',['../structdlaf_1_1_default_device.html',1,'dlaf']]], + ['defaultdevice_3c_20backend_3a_3agpu_20_3e_593',['DefaultDevice< Backend::GPU >',['../structdlaf_1_1_default_device_3_01_backend_1_1_g_p_u_01_4.html',1,'dlaf']]], + ['defaultdevice_3c_20backend_3a_3amc_20_3e_594',['DefaultDevice< Backend::MC >',['../structdlaf_1_1_default_device_3_01_backend_1_1_m_c_01_4.html',1,'dlaf']]], + ['distindexing_595',['DistIndexing',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_dist_indexing.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['distribution_596',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html',1,'dlaf::matrix']]], + ['distworkspacehostmirror_597',['DistWorkSpaceHostMirror',['../structdlaf_1_1eigensolver_1_1internal_1_1_dist_work_space_host_mirror.html',1,'dlaf::eigensolver::internal']]], + ['duplicatenocopy_598',['DuplicateNoCopy',['../structdlaf_1_1matrix_1_1_duplicate_no_copy.html',1,'dlaf::matrix']]] ]; diff --git a/master/search/classes_4.js b/master/search/classes_4.js index dcbc09f793..4fb65aeac4 100644 --- a/master/search/classes_4.js +++ b/master/search/classes_4.js @@ -1,6 +1,6 @@ var searchData= [ - ['eigensolver_595',['Eigensolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_eigensolver.html',1,'dlaf::eigensolver::internal']]], - ['eigensolverresult_596',['EigensolverResult',['../structdlaf_1_1eigensolver_1_1_eigensolver_result.html',1,'dlaf::eigensolver']]], - ['emptyenv_597',['EmptyEnv',['../structdlaf_1_1internal_1_1_empty_env.html',1,'dlaf::internal']]] + ['eigensolver_599',['Eigensolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_eigensolver.html',1,'dlaf::eigensolver::internal']]], + ['eigensolverresult_600',['EigensolverResult',['../structdlaf_1_1eigensolver_1_1_eigensolver_result.html',1,'dlaf::eigensolver']]], + ['emptyenv_601',['EmptyEnv',['../structdlaf_1_1internal_1_1_empty_env.html',1,'dlaf::internal']]] ]; diff --git a/master/search/classes_5.js b/master/search/classes_5.js index 20c3abc411..b5d05ebcf6 100644 --- a/master/search/classes_5.js +++ b/master/search/classes_5.js @@ -1,4 +1,4 @@ var searchData= [ - ['formatshort_598',['FormatShort',['../structdlaf_1_1internal_1_1_format_short.html',1,'dlaf::internal']]] + ['formatshort_602',['FormatShort',['../structdlaf_1_1internal_1_1_format_short.html',1,'dlaf::internal']]] ]; diff --git a/master/search/classes_6.js b/master/search/classes_6.js index bae08c3e91..f4042b192e 100644 --- a/master/search/classes_6.js +++ b/master/search/classes_6.js @@ -1,10 +1,10 @@ var searchData= [ - ['gemmsizes_599',['gemmSizes',['../structdlaf_1_1tile_1_1internal_1_1gemm_sizes.html',1,'dlaf::tile::internal']]], - ['geneigensolver_600',['GenEigensolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_gen_eigensolver.html',1,'dlaf::eigensolver::internal']]], - ['generalsub_601',['GeneralSub',['../structdlaf_1_1multiplication_1_1internal_1_1_general_sub.html',1,'dlaf::multiplication::internal']]], - ['gentostd_602',['GenToStd',['../structdlaf_1_1eigensolver_1_1internal_1_1_gen_to_std.html',1,'dlaf::eigensolver::internal']]], - ['getter_5frandom_603',['getter_random',['../classdlaf_1_1matrix_1_1util_1_1internal_1_1getter__random.html',1,'dlaf::matrix::util::internal']]], - ['getter_5frandom_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_604',['getter_random< std::complex< T > >',['../classdlaf_1_1matrix_1_1util_1_1internal_1_1getter__random_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::matrix::util::internal']]], - ['givensrotation_605',['GivensRotation',['../structdlaf_1_1eigensolver_1_1internal_1_1_givens_rotation.html',1,'dlaf::eigensolver::internal']]] + ['gemmsizes_603',['gemmSizes',['../structdlaf_1_1tile_1_1internal_1_1gemm_sizes.html',1,'dlaf::tile::internal']]], + ['geneigensolver_604',['GenEigensolver',['../structdlaf_1_1eigensolver_1_1internal_1_1_gen_eigensolver.html',1,'dlaf::eigensolver::internal']]], + ['generalsub_605',['GeneralSub',['../structdlaf_1_1multiplication_1_1internal_1_1_general_sub.html',1,'dlaf::multiplication::internal']]], + ['gentostd_606',['GenToStd',['../structdlaf_1_1eigensolver_1_1internal_1_1_gen_to_std.html',1,'dlaf::eigensolver::internal']]], + ['getter_5frandom_607',['getter_random',['../classdlaf_1_1matrix_1_1util_1_1internal_1_1getter__random.html',1,'dlaf::matrix::util::internal']]], + ['getter_5frandom_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_608',['getter_random< std::complex< T > >',['../classdlaf_1_1matrix_1_1util_1_1internal_1_1getter__random_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::matrix::util::internal']]], + ['givensrotation_609',['GivensRotation',['../structdlaf_1_1eigensolver_1_1internal_1_1_givens_rotation.html',1,'dlaf::eigensolver::internal']]] ]; diff --git a/master/search/classes_7.js b/master/search/classes_7.js index 57283c9ab7..0848569bec 100644 --- a/master/search/classes_7.js +++ b/master/search/classes_7.js @@ -1,15 +1,15 @@ var searchData= [ - ['helpers_606',['Helpers',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers.html',1,'dlaf::eigensolver::internal::bt_red_band::Helpers< B >'],['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers.html',1,'dlaf::factorization::internal::tfactor_l::Helpers< backend, device, T >']]], - ['helpers_3c_20backend_3a_3agpu_20_3e_607',['Helpers< Backend::GPU >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers_3_01_backend_1_1_g_p_u_01_4.html',1,'dlaf::eigensolver::internal::bt_red_band']]], - ['helpers_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_608',['Helpers< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers_3_01_backend_1_1_g_p_u_00_01_device_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::factorization::internal::tfactor_l']]], - ['helpers_3c_20backend_3a_3amc_20_3e_609',['Helpers< Backend::MC >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers_3_01_backend_1_1_m_c_01_4.html',1,'dlaf::eigensolver::internal::bt_red_band']]], - ['helpers_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_610',['Helpers< Backend::MC, Device::CPU, T >',['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::factorization::internal::tfactor_l']]], - ['hemmsizes_611',['hemmSizes',['../structdlaf_1_1tile_1_1internal_1_1hemm_sizes.html',1,'dlaf::tile::internal']]], - ['her2ksizes_612',['her2kSizes',['../structdlaf_1_1tile_1_1internal_1_1her2k_sizes.html',1,'dlaf::tile::internal']]], - ['herksizes_613',['herkSizes',['../structdlaf_1_1tile_1_1internal_1_1herk_sizes.html',1,'dlaf::tile::internal']]], - ['hermitian_614',['Hermitian',['../structdlaf_1_1multiplication_1_1internal_1_1_hermitian.html',1,'dlaf::multiplication::internal']]], - ['hhmanager_615',['HHManager',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['hhmanager_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_616',['HHManager< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager_3_01_backend_1_1_g_p_u_00_01_device_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], - ['hhmanager_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_617',['HHManager< Backend::MC, Device::CPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]] + ['helpers_610',['Helpers',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers.html',1,'dlaf::eigensolver::internal::bt_red_band::Helpers< B >'],['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers.html',1,'dlaf::factorization::internal::tfactor_l::Helpers< backend, device, T >']]], + ['helpers_3c_20backend_3a_3agpu_20_3e_611',['Helpers< Backend::GPU >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers_3_01_backend_1_1_g_p_u_01_4.html',1,'dlaf::eigensolver::internal::bt_red_band']]], + ['helpers_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_612',['Helpers< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers_3_01_backend_1_1_g_p_u_00_01_device_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::factorization::internal::tfactor_l']]], + ['helpers_3c_20backend_3a_3amc_20_3e_613',['Helpers< Backend::MC >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__red__band_1_1_helpers_3_01_backend_1_1_m_c_01_4.html',1,'dlaf::eigensolver::internal::bt_red_band']]], + ['helpers_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_614',['Helpers< Backend::MC, Device::CPU, T >',['../structdlaf_1_1factorization_1_1internal_1_1tfactor__l_1_1_helpers_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::factorization::internal::tfactor_l']]], + ['hemmsizes_615',['hemmSizes',['../structdlaf_1_1tile_1_1internal_1_1hemm_sizes.html',1,'dlaf::tile::internal']]], + ['her2ksizes_616',['her2kSizes',['../structdlaf_1_1tile_1_1internal_1_1her2k_sizes.html',1,'dlaf::tile::internal']]], + ['herksizes_617',['herkSizes',['../structdlaf_1_1tile_1_1internal_1_1herk_sizes.html',1,'dlaf::tile::internal']]], + ['hermitian_618',['Hermitian',['../structdlaf_1_1multiplication_1_1internal_1_1_hermitian.html',1,'dlaf::multiplication::internal']]], + ['hhmanager_619',['HHManager',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['hhmanager_3c_20backend_3a_3agpu_2c_20device_3a_3agpu_2c_20t_20_3e_620',['HHManager< Backend::GPU, Device::GPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager_3_01_backend_1_1_g_p_u_00_01_device_1_1_g_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]], + ['hhmanager_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_621',['HHManager< Backend::MC, Device::CPU, T >',['../structdlaf_1_1eigensolver_1_1internal_1_1bt__tridiag_1_1_h_h_manager_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::eigensolver::internal::bt_tridiag']]] ]; diff --git a/master/search/classes_8.js b/master/search/classes_8.js index 1ea701b7d2..052809a7da 100644 --- a/master/search/classes_8.js +++ b/master/search/classes_8.js @@ -1,18 +1,18 @@ var searchData= [ - ['index2d_618',['Index2D',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20indext_5fmpi_2c_20tag_5fmpi_20_3e_619',['Index2D< IndexT_MPI, TAG_MPI >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20sizetype_2c_20matrix_3a_3aglobalelement_5ftag_20_3e_620',['Index2D< SizeType, matrix::GlobalElement_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20sizetype_2c_20matrix_3a_3aglobaltile_5ftag_20_3e_621',['Index2D< SizeType, matrix::GlobalTile_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20sizetype_2c_20matrix_3a_3alocaltile_5ftag_20_3e_622',['Index2D< SizeType, matrix::LocalTile_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['index2d_3c_20sizetype_2c_20matrix_3a_3atileelement_5ftag_20_3e_623',['Index2D< SizeType, matrix::TileElement_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], - ['is_5fcoord_624',['is_coord',['../structdlaf_1_1common_1_1internal_1_1is__coord.html',1,'dlaf::common::internal']]], - ['is_5fcoord_3c_20index2d_3c_20t_2c_20tag_20_3e_20_3e_625',['is_coord< Index2D< T, Tag > >',['../structdlaf_1_1common_1_1internal_1_1is__coord_3_01_index2_d_3_01_t_00_01_tag_01_4_01_4.html',1,'dlaf::common::internal']]], - ['is_5fcoord_3c_20size2d_3c_20t_2c_20tag_20_3e_20_3e_626',['is_coord< Size2D< T, Tag > >',['../structdlaf_1_1common_1_1internal_1_1is__coord_3_01_size2_d_3_01_t_00_01_tag_01_4_01_4.html',1,'dlaf::common::internal']]], - ['is_5fdata_627',['is_data',['../structdlaf_1_1common_1_1is__data.html',1,'dlaf::common']]], - ['isfalse_628',['IsFalse',['../structdlaf_1_1internal_1_1_is_false.html',1,'dlaf::internal']]], - ['isfloatingpointorcomplex_629',['IsFloatingPointOrComplex',['../structdlaf_1_1internal_1_1_is_floating_point_or_complex.html',1,'dlaf::internal']]], - ['isfloatingpointorcomplex_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_630',['IsFloatingPointOrComplex< std::complex< T > >',['../structdlaf_1_1internal_1_1_is_floating_point_or_complex_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::internal']]], - ['iterablerange2d_631',['IterableRange2D',['../classdlaf_1_1common_1_1_iterable_range2_d.html',1,'dlaf::common']]], - ['iteratorrange2d_632',['IteratorRange2D',['../structdlaf_1_1common_1_1_iterator_range2_d.html',1,'dlaf::common']]] + ['index2d_622',['Index2D',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20indext_5fmpi_2c_20tag_5fmpi_20_3e_623',['Index2D< IndexT_MPI, TAG_MPI >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20sizetype_2c_20matrix_3a_3aglobalelement_5ftag_20_3e_624',['Index2D< SizeType, matrix::GlobalElement_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20sizetype_2c_20matrix_3a_3aglobaltile_5ftag_20_3e_625',['Index2D< SizeType, matrix::GlobalTile_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20sizetype_2c_20matrix_3a_3alocaltile_5ftag_20_3e_626',['Index2D< SizeType, matrix::LocalTile_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['index2d_3c_20sizetype_2c_20matrix_3a_3atileelement_5ftag_20_3e_627',['Index2D< SizeType, matrix::TileElement_TAG >',['../classdlaf_1_1common_1_1_index2_d.html',1,'dlaf::common']]], + ['is_5fcoord_628',['is_coord',['../structdlaf_1_1common_1_1internal_1_1is__coord.html',1,'dlaf::common::internal']]], + ['is_5fcoord_3c_20index2d_3c_20t_2c_20tag_20_3e_20_3e_629',['is_coord< Index2D< T, Tag > >',['../structdlaf_1_1common_1_1internal_1_1is__coord_3_01_index2_d_3_01_t_00_01_tag_01_4_01_4.html',1,'dlaf::common::internal']]], + ['is_5fcoord_3c_20size2d_3c_20t_2c_20tag_20_3e_20_3e_630',['is_coord< Size2D< T, Tag > >',['../structdlaf_1_1common_1_1internal_1_1is__coord_3_01_size2_d_3_01_t_00_01_tag_01_4_01_4.html',1,'dlaf::common::internal']]], + ['is_5fdata_631',['is_data',['../structdlaf_1_1common_1_1is__data.html',1,'dlaf::common']]], + ['isfalse_632',['IsFalse',['../structdlaf_1_1internal_1_1_is_false.html',1,'dlaf::internal']]], + ['isfloatingpointorcomplex_633',['IsFloatingPointOrComplex',['../structdlaf_1_1internal_1_1_is_floating_point_or_complex.html',1,'dlaf::internal']]], + ['isfloatingpointorcomplex_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_634',['IsFloatingPointOrComplex< std::complex< T > >',['../structdlaf_1_1internal_1_1_is_floating_point_or_complex_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::internal']]], + ['iterablerange2d_635',['IterableRange2D',['../classdlaf_1_1common_1_1_iterable_range2_d.html',1,'dlaf::common']]], + ['iteratorrange2d_636',['IteratorRange2D',['../structdlaf_1_1common_1_1_iterator_range2_d.html',1,'dlaf::common']]] ]; diff --git a/master/search/classes_9.js b/master/search/classes_9.js index 7f017c6104..73d9f3c522 100644 --- a/master/search/classes_9.js +++ b/master/search/classes_9.js @@ -1,4 +1,4 @@ var searchData= [ - ['layoutinfo_633',['LayoutInfo',['../classdlaf_1_1matrix_1_1_layout_info.html',1,'dlaf::matrix']]] + ['layoutinfo_637',['LayoutInfo',['../classdlaf_1_1matrix_1_1_layout_info.html',1,'dlaf::matrix']]] ]; diff --git a/master/search/classes_a.js b/master/search/classes_a.js index 374e3f19f8..61e07c08db 100644 --- a/master/search/classes_a.js +++ b/master/search/classes_a.js @@ -1,24 +1,24 @@ var searchData= [ - ['matrix_634',['Matrix',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20const_20t_2c_20d_20_3e_635',['Matrix< const T, D >',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html',1,'dlaf::matrix']]], - ['matrix_3c_20const_20t_2c_20source_20_3e_636',['Matrix< const T, Source >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20const_20t_2c_20sourcetarget_20_3e_637',['Matrix< const T, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20t_2c_20source_20_3e_638',['Matrix< T, Source >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20t_2c_20sourcetarget_20_3e_639',['Matrix< T, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrix_3c_20t_2c_20target_20_3e_640',['Matrix< T, Target >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], - ['matrixbase_641',['MatrixBase',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html',1,'dlaf::matrix::internal']]], - ['matrixmirror_642',['MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror.html',1,'dlaf::matrix']]], - ['matrixmirror_3c_20const_20t_2c_20sourcetarget_2c_20sourcetarget_20_3e_643',['MatrixMirror< const T, SourceTarget, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html',1,'dlaf::matrix']]], - ['matrixmirror_3c_20const_20t_2c_20target_2c_20source_20_3e_644',['MatrixMirror< const T, Target, Source >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html',1,'dlaf::matrix']]], - ['matrixmirror_3c_20t_2c_20sourcetarget_2c_20sourcetarget_20_3e_645',['MatrixMirror< T, SourceTarget, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html',1,'dlaf::matrix']]], - ['memorychunk_646',['MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html',1,'dlaf::memory']]], - ['memoryview_647',['MemoryView',['../classdlaf_1_1memory_1_1_memory_view.html',1,'dlaf::memory']]], - ['memoryview_3c_20t_2c_20device_3a_3acpu_20_3e_648',['MemoryView< T, Device::CPU >',['../classdlaf_1_1memory_1_1_memory_view.html',1,'dlaf::memory']]], - ['message_649',['Message',['../classdlaf_1_1comm_1_1_message.html',1,'dlaf::comm']]], - ['movenonconsttile_650',['moveNonConstTile',['../structdlaf_1_1internal_1_1move_non_const_tile.html',1,'dlaf::internal']]], - ['mpi_5fdatatype_651',['mpi_datatype',['../structdlaf_1_1comm_1_1mpi__datatype.html',1,'dlaf::comm']]], - ['mpi_5fdatatype_3c_20const_20t_20_3e_652',['mpi_datatype< const T >',['../structdlaf_1_1comm_1_1mpi__datatype_3_01const_01_t_01_4.html',1,'dlaf::comm']]], - ['mpi_5finit_653',['mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html',1,'dlaf::comm']]], - ['mpicallhelper_654',['MPICallHelper',['../structdlaf_1_1comm_1_1internal_1_1_m_p_i_call_helper.html',1,'dlaf::comm::internal']]] + ['matrix_638',['Matrix',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20const_20t_2c_20d_20_3e_639',['Matrix< const T, D >',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html',1,'dlaf::matrix']]], + ['matrix_3c_20const_20t_2c_20source_20_3e_640',['Matrix< const T, Source >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20const_20t_2c_20sourcetarget_20_3e_641',['Matrix< const T, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20t_2c_20source_20_3e_642',['Matrix< T, Source >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20t_2c_20sourcetarget_20_3e_643',['Matrix< T, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrix_3c_20t_2c_20target_20_3e_644',['Matrix< T, Target >',['../classdlaf_1_1matrix_1_1_matrix.html',1,'dlaf::matrix']]], + ['matrixbase_645',['MatrixBase',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html',1,'dlaf::matrix::internal']]], + ['matrixmirror_646',['MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror.html',1,'dlaf::matrix']]], + ['matrixmirror_3c_20const_20t_2c_20sourcetarget_2c_20sourcetarget_20_3e_647',['MatrixMirror< const T, SourceTarget, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html',1,'dlaf::matrix']]], + ['matrixmirror_3c_20const_20t_2c_20target_2c_20source_20_3e_648',['MatrixMirror< const T, Target, Source >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html',1,'dlaf::matrix']]], + ['matrixmirror_3c_20t_2c_20sourcetarget_2c_20sourcetarget_20_3e_649',['MatrixMirror< T, SourceTarget, SourceTarget >',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html',1,'dlaf::matrix']]], + ['memorychunk_650',['MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html',1,'dlaf::memory']]], + ['memoryview_651',['MemoryView',['../classdlaf_1_1memory_1_1_memory_view.html',1,'dlaf::memory']]], + ['memoryview_3c_20t_2c_20device_3a_3acpu_20_3e_652',['MemoryView< T, Device::CPU >',['../classdlaf_1_1memory_1_1_memory_view.html',1,'dlaf::memory']]], + ['message_653',['Message',['../classdlaf_1_1comm_1_1_message.html',1,'dlaf::comm']]], + ['movenonconsttile_654',['moveNonConstTile',['../structdlaf_1_1internal_1_1move_non_const_tile.html',1,'dlaf::internal']]], + ['mpi_5fdatatype_655',['mpi_datatype',['../structdlaf_1_1comm_1_1mpi__datatype.html',1,'dlaf::comm']]], + ['mpi_5fdatatype_3c_20const_20t_20_3e_656',['mpi_datatype< const T >',['../structdlaf_1_1comm_1_1mpi__datatype_3_01const_01_t_01_4.html',1,'dlaf::comm']]], + ['mpi_5finit_657',['mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html',1,'dlaf::comm']]], + ['mpicallhelper_658',['MPICallHelper',['../structdlaf_1_1comm_1_1internal_1_1_m_p_i_call_helper.html',1,'dlaf::comm::internal']]] ]; diff --git a/master/search/classes_b.js b/master/search/classes_b.js index d76b7f0ab9..3e07ead1d1 100644 --- a/master/search/classes_b.js +++ b/master/search/classes_b.js @@ -1,8 +1,8 @@ var searchData= [ - ['norm_655',['Norm',['../structdlaf_1_1auxiliary_1_1internal_1_1_norm.html',1,'dlaf::auxiliary::internal']]], - ['norm_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_656',['Norm< Backend::MC, Device::CPU, T >',['../structdlaf_1_1auxiliary_1_1internal_1_1_norm_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::auxiliary::internal']]], - ['numpy_657',['numpy',['../structdlaf_1_1format_1_1numpy.html',1,'dlaf::format']]], - ['numpy_5fdatatype_658',['numpy_datatype',['../structdlaf_1_1matrix_1_1internal_1_1numpy__datatype.html',1,'dlaf::matrix::internal']]], - ['numpy_5fdatatype_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_659',['numpy_datatype< std::complex< T > >',['../structdlaf_1_1matrix_1_1internal_1_1numpy__datatype_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::matrix::internal']]] + ['norm_659',['Norm',['../structdlaf_1_1auxiliary_1_1internal_1_1_norm.html',1,'dlaf::auxiliary::internal']]], + ['norm_3c_20backend_3a_3amc_2c_20device_3a_3acpu_2c_20t_20_3e_660',['Norm< Backend::MC, Device::CPU, T >',['../structdlaf_1_1auxiliary_1_1internal_1_1_norm_3_01_backend_1_1_m_c_00_01_device_1_1_c_p_u_00_01_t_01_4.html',1,'dlaf::auxiliary::internal']]], + ['numpy_661',['numpy',['../structdlaf_1_1format_1_1numpy.html',1,'dlaf::format']]], + ['numpy_5fdatatype_662',['numpy_datatype',['../structdlaf_1_1matrix_1_1internal_1_1numpy__datatype.html',1,'dlaf::matrix::internal']]], + ['numpy_5fdatatype_3c_20std_3a_3acomplex_3c_20t_20_3e_20_3e_663',['numpy_datatype< std::complex< T > >',['../structdlaf_1_1matrix_1_1internal_1_1numpy__datatype_3_01std_1_1complex_3_01_t_01_4_01_4.html',1,'dlaf::matrix::internal']]] ]; diff --git a/master/search/classes_c.js b/master/search/classes_c.js index 4cded67c7d..113eb3e39d 100644 --- a/master/search/classes_c.js +++ b/master/search/classes_c.js @@ -1,18 +1,18 @@ var searchData= [ - ['panel_660',['Panel',['../structdlaf_1_1matrix_1_1_panel.html',1,'dlaf::matrix']]], - ['panel_3c_20axis_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ano_20_3e_661',['Panel< axis, const T, D, StoreTransposed::No >',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html',1,'dlaf::matrix']]], - ['panel_3c_20axis_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ayes_20_3e_662',['Panel< axis, const T, D, StoreTransposed::Yes >',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_yes_01_4.html',1,'dlaf::matrix']]], - ['panel_3c_20orthogonal_28axis_29_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ano_20_3e_663',['Panel< orthogonal(axis), const T, D, StoreTransposed::No >',['../structdlaf_1_1matrix_1_1_panel.html',1,'dlaf::matrix']]], - ['partialtransform_664',['PartialTransform',['../classdlaf_1_1internal_1_1_partial_transform.html',1,'dlaf::internal']]], - ['partialtransformbase_665',['PartialTransformBase',['../structdlaf_1_1internal_1_1_partial_transform_base.html',1,'dlaf::internal']]], - ['partialtransformdetach_666',['PartialTransformDetach',['../classdlaf_1_1internal_1_1_partial_transform_detach.html',1,'dlaf::internal']]], - ['partialtransformmpi_667',['PartialTransformMPI',['../classdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i.html',1,'dlaf::comm::internal']]], - ['partialtransformmpibase_668',['PartialTransformMPIBase',['../structdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i_base.html',1,'dlaf::comm::internal']]], - ['partialtransformmpidetach_669',['PartialTransformMPIDetach',['../classdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i_detach.html',1,'dlaf::comm::internal']]], - ['permutations_670',['Permutations',['../structdlaf_1_1permutations_1_1internal_1_1_permutations.html',1,'dlaf::permutations::internal']]], - ['pipeline_671',['Pipeline',['../classdlaf_1_1common_1_1_pipeline.html',1,'dlaf::common']]], - ['policy_672',['Policy',['../classdlaf_1_1internal_1_1_policy.html',1,'dlaf::internal']]], - ['prependpack_673',['PrependPack',['../structdlaf_1_1internal_1_1_prepend_pack.html',1,'dlaf::internal']]], - ['prependpack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20t_20_3e_674',['PrependPack< Pack< Ts... >, T >',['../structdlaf_1_1internal_1_1_prepend_pack_3_01_pack_3_01_ts_8_8_8_01_4_00_01_t_01_4.html',1,'dlaf::internal']]] + ['panel_664',['Panel',['../structdlaf_1_1matrix_1_1_panel.html',1,'dlaf::matrix']]], + ['panel_3c_20axis_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ano_20_3e_665',['Panel< axis, const T, D, StoreTransposed::No >',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html',1,'dlaf::matrix']]], + ['panel_3c_20axis_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ayes_20_3e_666',['Panel< axis, const T, D, StoreTransposed::Yes >',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_yes_01_4.html',1,'dlaf::matrix']]], + ['panel_3c_20orthogonal_28axis_29_2c_20const_20t_2c_20d_2c_20storetransposed_3a_3ano_20_3e_667',['Panel< orthogonal(axis), const T, D, StoreTransposed::No >',['../structdlaf_1_1matrix_1_1_panel.html',1,'dlaf::matrix']]], + ['partialtransform_668',['PartialTransform',['../classdlaf_1_1internal_1_1_partial_transform.html',1,'dlaf::internal']]], + ['partialtransformbase_669',['PartialTransformBase',['../structdlaf_1_1internal_1_1_partial_transform_base.html',1,'dlaf::internal']]], + ['partialtransformdetach_670',['PartialTransformDetach',['../classdlaf_1_1internal_1_1_partial_transform_detach.html',1,'dlaf::internal']]], + ['partialtransformmpi_671',['PartialTransformMPI',['../classdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i.html',1,'dlaf::comm::internal']]], + ['partialtransformmpibase_672',['PartialTransformMPIBase',['../structdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i_base.html',1,'dlaf::comm::internal']]], + ['partialtransformmpidetach_673',['PartialTransformMPIDetach',['../classdlaf_1_1comm_1_1internal_1_1_partial_transform_m_p_i_detach.html',1,'dlaf::comm::internal']]], + ['permutations_674',['Permutations',['../structdlaf_1_1permutations_1_1internal_1_1_permutations.html',1,'dlaf::permutations::internal']]], + ['pipeline_675',['Pipeline',['../classdlaf_1_1common_1_1_pipeline.html',1,'dlaf::common']]], + ['policy_676',['Policy',['../classdlaf_1_1internal_1_1_policy.html',1,'dlaf::internal']]], + ['prependpack_677',['PrependPack',['../structdlaf_1_1internal_1_1_prepend_pack.html',1,'dlaf::internal']]], + ['prependpack_3c_20pack_3c_20ts_2e_2e_2e_20_3e_2c_20t_20_3e_678',['PrependPack< Pack< Ts... >, T >',['../structdlaf_1_1internal_1_1_prepend_pack_3_01_pack_3_01_ts_8_8_8_01_4_00_01_t_01_4.html',1,'dlaf::internal']]] ]; diff --git a/master/search/classes_d.js b/master/search/classes_d.js index e4f333b0c2..9d11f696a8 100644 --- a/master/search/classes_d.js +++ b/master/search/classes_d.js @@ -1,5 +1,5 @@ var searchData= [ - ['qr_675',['QR',['../structdlaf_1_1factorization_1_1internal_1_1_q_r.html',1,'dlaf::factorization::internal']]], - ['qr_5ftfactor_676',['QR_Tfactor',['../structdlaf_1_1factorization_1_1internal_1_1_q_r___tfactor.html',1,'dlaf::factorization::internal']]] + ['qr_679',['QR',['../structdlaf_1_1factorization_1_1internal_1_1_q_r.html',1,'dlaf::factorization::internal']]], + ['qr_5ftfactor_680',['QR_Tfactor',['../structdlaf_1_1factorization_1_1internal_1_1_q_r___tfactor.html',1,'dlaf::factorization::internal']]] ]; diff --git a/master/search/classes_e.js b/master/search/classes_e.js index fc95f63c2f..42e733369c 100644 --- a/master/search/classes_e.js +++ b/master/search/classes_e.js @@ -1,7 +1,6 @@ var searchData= [ - ['reductiontoband_677',['ReductionToBand',['../structdlaf_1_1eigensolver_1_1internal_1_1_reduction_to_band.html',1,'dlaf::eigensolver::internal']]], - ['retiledmatrix_678',['RetiledMatrix',['../classdlaf_1_1matrix_1_1_retiled_matrix.html',1,'dlaf::matrix']]], - ['roundrobin_679',['RoundRobin',['../structdlaf_1_1common_1_1_round_robin.html',1,'dlaf::common']]], - ['roundrobin_3c_20dlaf_3a_3amatrix_3a_3apanel_3c_20coord_3a_3acol_2c_20t_2c_20device_3a_3acpu_20_3e_20_3e_680',['RoundRobin< dlaf::matrix::Panel< Coord::Col, T, Device::CPU > >',['../structdlaf_1_1common_1_1_round_robin.html',1,'dlaf::common']]] + ['reductiontoband_681',['ReductionToBand',['../structdlaf_1_1eigensolver_1_1internal_1_1_reduction_to_band.html',1,'dlaf::eigensolver::internal']]], + ['roundrobin_682',['RoundRobin',['../structdlaf_1_1common_1_1_round_robin.html',1,'dlaf::common']]], + ['roundrobin_3c_20dlaf_3a_3amatrix_3a_3apanel_3c_20coord_3a_3acol_2c_20t_2c_20device_3a_3acpu_20_3e_20_3e_683',['RoundRobin< dlaf::matrix::Panel< Coord::Col, T, Device::CPU > >',['../structdlaf_1_1common_1_1_round_robin.html',1,'dlaf::common']]] ]; diff --git a/master/search/classes_f.js b/master/search/classes_f.js index f501c6afea..752e9e24c6 100644 --- a/master/search/classes_f.js +++ b/master/search/classes_f.js @@ -1,24 +1,24 @@ var searchData= [ - ['scopedinitializer_681',['ScopedInitializer',['../structdlaf_1_1_scoped_initializer.html',1,'dlaf']]], - ['sendersinglevaluetypeimpl_682',['SenderSingleValueTypeImpl',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl.html',1,'dlaf::internal']]], - ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20rwtype_2c_20rtype_2c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5ftype_3a_3aread_20_3e_20_3e_20_3e_20_3e_683',['SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::read > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01pika_356df456d49829a554002eacb85cc5b7.html',1,'dlaf::internal']]], - ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20rwtype_2c_20rtype_2c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5ftype_3a_3areadwrite_20_3e_20_3e_20_3e_20_3e_684',['SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::readwrite > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01pika_3535bbe9d77284105a6158ec315ee38c.html',1,'dlaf::internal']]], - ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20std_3a_3areference_5fwrapper_3c_20t_20_3e_20_3e_20_3e_20_3e_685',['SenderSingleValueTypeImpl< TypeList< TypeList< std::reference_wrapper< T > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01std_1b569951bd408b620bb545799a6089d25.html',1,'dlaf::internal']]], - ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20t_20_3e_20_3e_20_3e_686',['SenderSingleValueTypeImpl< TypeList< TypeList< T > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01_t_01_4_01_4_01_4.html',1,'dlaf::internal']]], - ['singlethreadedblasscope_687',['SingleThreadedBlasScope',['../classdlaf_1_1common_1_1internal_1_1_single_threaded_blas_scope.html',1,'dlaf::common::internal']]], - ['size2d_688',['Size2D',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20indext_5fmpi_2c_20tag_5fmpi_20_3e_689',['Size2D< IndexT_MPI, TAG_MPI >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3aglobalelement_5ftag_20_3e_690',['Size2D< SizeType, matrix::GlobalElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3aglobaltile_5ftag_20_3e_691',['Size2D< SizeType, matrix::GlobalTile_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3alocalelement_5ftag_20_3e_692',['Size2D< SizeType, matrix::LocalElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3alocaltile_5ftag_20_3e_693',['Size2D< SizeType, matrix::LocalTile_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['size2d_3c_20sizetype_2c_20matrix_3a_3atileelement_5ftag_20_3e_694',['Size2D< SizeType, matrix::TileElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], - ['source_5flocation_695',['source_location',['../structdlaf_1_1common_1_1internal_1_1source__location.html',1,'dlaf::common::internal']]], - ['submatrixview_696',['SubMatrixView',['../structdlaf_1_1matrix_1_1_sub_matrix_view.html',1,'dlaf::matrix']]], - ['subpanelview_697',['SubPanelView',['../structdlaf_1_1matrix_1_1_sub_panel_view.html',1,'dlaf::matrix']]], - ['subpipelinetag_698',['SubPipelineTag',['../structdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4_1_1_sub_pipeline_tag.html',1,'dlaf::matrix::Matrix< const T, D >']]], - ['subtilespec_699',['SubTileSpec',['../structdlaf_1_1matrix_1_1_sub_tile_spec.html',1,'dlaf::matrix']]], - ['sweepworker_700',['SweepWorker',['../classdlaf_1_1eigensolver_1_1internal_1_1_sweep_worker.html',1,'dlaf::eigensolver::internal']]], - ['sweepworkerdist_701',['SweepWorkerDist',['../classdlaf_1_1eigensolver_1_1internal_1_1_sweep_worker_dist.html',1,'dlaf::eigensolver::internal']]] + ['scopedinitializer_684',['ScopedInitializer',['../structdlaf_1_1_scoped_initializer.html',1,'dlaf']]], + ['sendersinglevaluetypeimpl_685',['SenderSingleValueTypeImpl',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl.html',1,'dlaf::internal']]], + ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20rwtype_2c_20rtype_2c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5ftype_3a_3aread_20_3e_20_3e_20_3e_20_3e_686',['SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::read > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01pika_356df456d49829a554002eacb85cc5b7.html',1,'dlaf::internal']]], + ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5fwrapper_3c_20rwtype_2c_20rtype_2c_20pika_3a_3aexecution_3a_3aexperimental_3a_3aasync_5frw_5fmutex_5faccess_5ftype_3a_3areadwrite_20_3e_20_3e_20_3e_20_3e_687',['SenderSingleValueTypeImpl< TypeList< TypeList< pika::execution::experimental::async_rw_mutex_access_wrapper< RWType, RType, pika::execution::experimental::async_rw_mutex_access_type::readwrite > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01pika_3535bbe9d77284105a6158ec315ee38c.html',1,'dlaf::internal']]], + ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20std_3a_3areference_5fwrapper_3c_20t_20_3e_20_3e_20_3e_20_3e_688',['SenderSingleValueTypeImpl< TypeList< TypeList< std::reference_wrapper< T > > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01std_1b569951bd408b620bb545799a6089d25.html',1,'dlaf::internal']]], + ['sendersinglevaluetypeimpl_3c_20typelist_3c_20typelist_3c_20t_20_3e_20_3e_20_3e_689',['SenderSingleValueTypeImpl< TypeList< TypeList< T > > >',['../structdlaf_1_1internal_1_1_sender_single_value_type_impl_3_01_type_list_3_01_type_list_3_01_t_01_4_01_4_01_4.html',1,'dlaf::internal']]], + ['singlethreadedblasscope_690',['SingleThreadedBlasScope',['../classdlaf_1_1common_1_1internal_1_1_single_threaded_blas_scope.html',1,'dlaf::common::internal']]], + ['size2d_691',['Size2D',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20indext_5fmpi_2c_20tag_5fmpi_20_3e_692',['Size2D< IndexT_MPI, TAG_MPI >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3aglobalelement_5ftag_20_3e_693',['Size2D< SizeType, matrix::GlobalElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3aglobaltile_5ftag_20_3e_694',['Size2D< SizeType, matrix::GlobalTile_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3alocalelement_5ftag_20_3e_695',['Size2D< SizeType, matrix::LocalElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3alocaltile_5ftag_20_3e_696',['Size2D< SizeType, matrix::LocalTile_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['size2d_3c_20sizetype_2c_20matrix_3a_3atileelement_5ftag_20_3e_697',['Size2D< SizeType, matrix::TileElement_TAG >',['../classdlaf_1_1common_1_1_size2_d.html',1,'dlaf::common']]], + ['source_5flocation_698',['source_location',['../structdlaf_1_1common_1_1internal_1_1source__location.html',1,'dlaf::common::internal']]], + ['submatrixview_699',['SubMatrixView',['../structdlaf_1_1matrix_1_1_sub_matrix_view.html',1,'dlaf::matrix']]], + ['subpanelview_700',['SubPanelView',['../structdlaf_1_1matrix_1_1_sub_panel_view.html',1,'dlaf::matrix']]], + ['subpipelinetag_701',['SubPipelineTag',['../structdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4_1_1_sub_pipeline_tag.html',1,'dlaf::matrix::Matrix< const T, D >']]], + ['subtilespec_702',['SubTileSpec',['../structdlaf_1_1matrix_1_1_sub_tile_spec.html',1,'dlaf::matrix']]], + ['sweepworker_703',['SweepWorker',['../classdlaf_1_1eigensolver_1_1internal_1_1_sweep_worker.html',1,'dlaf::eigensolver::internal']]], + ['sweepworkerdist_704',['SweepWorkerDist',['../classdlaf_1_1eigensolver_1_1internal_1_1_sweep_worker_dist.html',1,'dlaf::eigensolver::internal']]] ]; diff --git a/master/search/defines_0.js b/master/search/defines_0.js index fad889f1ec..1ef5bcd9e6 100644 --- a/master/search/defines_0.js +++ b/master/search/defines_0.js @@ -1,4 +1,4 @@ var searchData= [ - ['dlaf_5fmake_5fcallable_5fobject_1088',['DLAF_MAKE_CALLABLE_OBJECT',['../callable__object_8h.html#a5aafa5a9c65ad07da48410427f4825b2',1,'callable_object.h']]] + ['dlaf_5fmake_5fcallable_5fobject_1094',['DLAF_MAKE_CALLABLE_OBJECT',['../callable__object_8h.html#a5aafa5a9c65ad07da48410427f4825b2',1,'callable_object.h']]] ]; diff --git a/master/search/defines_1.js b/master/search/defines_1.js index 6e19ff6ae1..ab45cfe071 100644 --- a/master/search/defines_1.js +++ b/master/search/defines_1.js @@ -1,4 +1,4 @@ var searchData= [ - ['source_5flocation_1089',['SOURCE_LOCATION',['../source__location_8h.html#aa6e6fc9684bd84ce7b72da6a2eef668f',1,'source_location.h']]] + ['source_5flocation_1095',['SOURCE_LOCATION',['../source__location_8h.html#aa6e6fc9684bd84ce7b72da6a2eef668f',1,'source_location.h']]] ]; diff --git a/master/search/enums_0.js b/master/search/enums_0.js index 7d6f99ad27..1909209e7e 100644 --- a/master/search/enums_0.js +++ b/master/search/enums_0.js @@ -1,4 +1,4 @@ var searchData= [ - ['ordering_1084',['Ordering',['../index2d_8h.html#a02419996f0ffc187079c310f0175ce76',1,'dlaf::common']]] + ['ordering_1090',['Ordering',['../index2d_8h.html#a02419996f0ffc187079c310f0175ce76',1,'dlaf::common']]] ]; diff --git a/master/search/files_0.js b/master/search/files_0.js index a3751432a0..257f497888 100644 --- a/master/search/files_0.js +++ b/master/search/files_0.js @@ -1,5 +1,5 @@ var searchData= [ - ['all_5freduce_2eh_745',['all_reduce.h',['../kernels_2all__reduce_8h.html',1,'(Global Namespace)'],['../sync_2all__reduce_8h.html',1,'(Global Namespace)']]], - ['auxiliary_2eh_746',['auxiliary.h',['../auxiliary_8h.html',1,'']]] + ['all_5freduce_2eh_748',['all_reduce.h',['../kernels_2all__reduce_8h.html',1,'(Global Namespace)'],['../sync_2all__reduce_8h.html',1,'(Global Namespace)']]], + ['auxiliary_2eh_749',['auxiliary.h',['../auxiliary_8h.html',1,'']]] ]; diff --git a/master/search/files_1.js b/master/search/files_1.js index 9995c9de67..fa7a870b18 100644 --- a/master/search/files_1.js +++ b/master/search/files_1.js @@ -1,9 +1,9 @@ var searchData= [ - ['band_5fto_5ftridiag_2eh_747',['band_to_tridiag.h',['../band__to__tridiag_8h.html',1,'']]], - ['basic_2eh_748',['basic.h',['../basic_8h.html',1,'']]], - ['broadcast_2eh_749',['broadcast.h',['../kernels_2broadcast_8h.html',1,'(Global Namespace)'],['../sync_2broadcast_8h.html',1,'(Global Namespace)']]], - ['broadcast_5fpanel_2eh_750',['broadcast_panel.h',['../broadcast__panel_8h.html',1,'']]], - ['bt_5fband_5fto_5ftridiag_2eh_751',['bt_band_to_tridiag.h',['../bt__band__to__tridiag_8h.html',1,'']]], - ['bt_5freduction_5fto_5fband_2eh_752',['bt_reduction_to_band.h',['../bt__reduction__to__band_8h.html',1,'']]] + ['band_5fto_5ftridiag_2eh_750',['band_to_tridiag.h',['../band__to__tridiag_8h.html',1,'']]], + ['basic_2eh_751',['basic.h',['../basic_8h.html',1,'']]], + ['broadcast_2eh_752',['broadcast.h',['../kernels_2broadcast_8h.html',1,'(Global Namespace)'],['../sync_2broadcast_8h.html',1,'(Global Namespace)']]], + ['broadcast_5fpanel_2eh_753',['broadcast_panel.h',['../broadcast__panel_8h.html',1,'']]], + ['bt_5fband_5fto_5ftridiag_2eh_754',['bt_band_to_tridiag.h',['../bt__band__to__tridiag_8h.html',1,'']]], + ['bt_5freduction_5fto_5fband_2eh_755',['bt_reduction_to_band.h',['../bt__reduction__to__band_8h.html',1,'']]] ]; diff --git a/master/search/files_10.js b/master/search/files_10.js index 85b864e224..0b346cc6e0 100644 --- a/master/search/files_10.js +++ b/master/search/files_10.js @@ -1,8 +1,8 @@ var searchData= [ - ['scal_2eh_799',['scal.h',['../scal_8h.html',1,'']]], - ['schedulers_2eh_800',['schedulers.h',['../schedulers_8h.html',1,'']]], - ['single_5fthreaded_5fblas_2eh_801',['single_threaded_blas.h',['../single__threaded__blas_8h.html',1,'']]], - ['solver_2eh_802',['solver.h',['../solver_8h.html',1,'']]], - ['source_5flocation_2eh_803',['source_location.h',['../source__location_8h.html',1,'']]] + ['scal_2eh_801',['scal.h',['../scal_8h.html',1,'']]], + ['schedulers_2eh_802',['schedulers.h',['../schedulers_8h.html',1,'']]], + ['single_5fthreaded_5fblas_2eh_803',['single_threaded_blas.h',['../single__threaded__blas_8h.html',1,'']]], + ['solver_2eh_804',['solver.h',['../solver_8h.html',1,'']]], + ['source_5flocation_2eh_805',['source_location.h',['../source__location_8h.html',1,'']]] ]; diff --git a/master/search/files_11.js b/master/search/files_11.js index d9dd35f71b..d70fa6521c 100644 --- a/master/search/files_11.js +++ b/master/search/files_11.js @@ -1,12 +1,12 @@ var searchData= [ - ['tile_2eh_804',['tile.h',['../blas_2tile_8h.html',1,'(Global Namespace)'],['../lapack_2tile_8h.html',1,'(Global Namespace)'],['../matrix_2tile_8h.html',1,'(Global Namespace)']]], - ['tile_5fextensions_2eh_805',['tile_extensions.h',['../tile__extensions_8h.html',1,'']]], - ['timer_2eh_806',['timer.h',['../timer_8h.html',1,'']]], - ['traits_2eh_807',['traits.h',['../traits_8h.html',1,'']]], - ['triangular_2eh_808',['triangular.h',['../multiplication_2triangular_8h.html',1,'(Global Namespace)'],['../solver_2triangular_8h.html',1,'(Global Namespace)']]], - ['tridiag_5fsolver_2eh_809',['tridiag_solver.h',['../tridiag__solver_8h.html',1,'']]], - ['tune_2eh_810',['tune.h',['../tune_8h.html',1,'']]], - ['type_5fhandler_2eh_811',['type_handler.h',['../type__handler_8h.html',1,'']]], - ['types_2eh_812',['types.h',['../types_8h.html',1,'']]] + ['tile_2eh_806',['tile.h',['../blas_2tile_8h.html',1,'(Global Namespace)'],['../lapack_2tile_8h.html',1,'(Global Namespace)'],['../matrix_2tile_8h.html',1,'(Global Namespace)']]], + ['tile_5fextensions_2eh_807',['tile_extensions.h',['../tile__extensions_8h.html',1,'']]], + ['timer_2eh_808',['timer.h',['../timer_8h.html',1,'']]], + ['traits_2eh_809',['traits.h',['../traits_8h.html',1,'']]], + ['triangular_2eh_810',['triangular.h',['../multiplication_2triangular_8h.html',1,'(Global Namespace)'],['../solver_2triangular_8h.html',1,'(Global Namespace)']]], + ['tridiag_5fsolver_2eh_811',['tridiag_solver.h',['../tridiag__solver_8h.html',1,'']]], + ['tune_2eh_812',['tune.h',['../tune_8h.html',1,'']]], + ['type_5fhandler_2eh_813',['type_handler.h',['../type__handler_8h.html',1,'']]], + ['types_2eh_814',['types.h',['../types_8h.html',1,'']]] ]; diff --git a/master/search/files_12.js b/master/search/files_12.js index 621bef8b11..a66b109a99 100644 --- a/master/search/files_12.js +++ b/master/search/files_12.js @@ -1,10 +1,10 @@ var searchData= [ - ['unwrap_2eh_813',['unwrap.h',['../unwrap_8h.html',1,'']]], - ['util_5fblas_2eh_814',['util_blas.h',['../util__blas_8h.html',1,'']]], - ['util_5fdistribution_2eh_815',['util_distribution.h',['../util__distribution_8h.html',1,'']]], - ['util_5flapack_2eh_816',['util_lapack.h',['../util__lapack_8h.html',1,'']]], - ['util_5fmath_2eh_817',['util_math.h',['../util__math_8h.html',1,'']]], - ['util_5fmatrix_2eh_818',['util_matrix.h',['../util__matrix_8h.html',1,'']]], - ['util_5ftile_2eh_819',['util_tile.h',['../util__tile_8h.html',1,'']]] + ['unwrap_2eh_815',['unwrap.h',['../unwrap_8h.html',1,'']]], + ['util_5fblas_2eh_816',['util_blas.h',['../util__blas_8h.html',1,'']]], + ['util_5fdistribution_2eh_817',['util_distribution.h',['../util__distribution_8h.html',1,'']]], + ['util_5flapack_2eh_818',['util_lapack.h',['../util__lapack_8h.html',1,'']]], + ['util_5fmath_2eh_819',['util_math.h',['../util__math_8h.html',1,'']]], + ['util_5fmatrix_2eh_820',['util_matrix.h',['../util__matrix_8h.html',1,'']]], + ['util_5ftile_2eh_821',['util_tile.h',['../util__tile_8h.html',1,'']]] ]; diff --git a/master/search/files_13.js b/master/search/files_13.js index 798f9bda85..cd1a0c6b80 100644 --- a/master/search/files_13.js +++ b/master/search/files_13.js @@ -1,5 +1,5 @@ var searchData= [ - ['vector_2eh_820',['vector.h',['../vector_8h.html',1,'']]], - ['views_2eh_821',['views.h',['../views_8h.html',1,'']]] + ['vector_2eh_822',['vector.h',['../vector_8h.html',1,'']]], + ['views_2eh_823',['views.h',['../views_8h.html',1,'']]] ]; diff --git a/master/search/files_2.js b/master/search/files_2.js index 816d3cefbb..d44a0d2e82 100644 --- a/master/search/files_2.js +++ b/master/search/files_2.js @@ -1,9 +1,9 @@ var searchData= [ - ['callable_5fobject_2eh_753',['callable_object.h',['../callable__object_8h.html',1,'']]], - ['cholesky_2eh_754',['cholesky.h',['../cholesky_8h.html',1,'']]], - ['communicator_2eh_755',['communicator.h',['../communicator_8h.html',1,'']]], - ['communicator_5fgrid_2eh_756',['communicator_grid.h',['../communicator__grid_8h.html',1,'']]], - ['copy_2eh_757',['copy.h',['../copy_8h.html',1,'']]], - ['copy_5ftile_2eh_758',['copy_tile.h',['../copy__tile_8h.html',1,'']]] + ['callable_5fobject_2eh_756',['callable_object.h',['../callable__object_8h.html',1,'']]], + ['cholesky_2eh_757',['cholesky.h',['../cholesky_8h.html',1,'']]], + ['communicator_2eh_758',['communicator.h',['../communicator_8h.html',1,'']]], + ['communicator_5fgrid_2eh_759',['communicator_grid.h',['../communicator__grid_8h.html',1,'']]], + ['copy_2eh_760',['copy.h',['../copy_8h.html',1,'']]], + ['copy_5ftile_2eh_761',['copy_tile.h',['../copy__tile_8h.html',1,'']]] ]; diff --git a/master/search/files_3.js b/master/search/files_3.js index f3818d6f25..7bc5f78f93 100644 --- a/master/search/files_3.js +++ b/master/search/files_3.js @@ -1,7 +1,7 @@ var searchData= [ - ['data_2eh_759',['data.h',['../data_8h.html',1,'']]], - ['data_5fdescriptor_2eh_760',['data_descriptor.h',['../data__descriptor_8h.html',1,'']]], - ['datatypes_2eh_761',['datatypes.h',['../datatypes_8h.html',1,'']]], - ['distribution_2eh_762',['distribution.h',['../distribution_8h.html',1,'']]] + ['data_2eh_762',['data.h',['../data_8h.html',1,'']]], + ['data_5fdescriptor_2eh_763',['data_descriptor.h',['../data__descriptor_8h.html',1,'']]], + ['datatypes_2eh_764',['datatypes.h',['../datatypes_8h.html',1,'']]], + ['distribution_2eh_765',['distribution.h',['../distribution_8h.html',1,'']]] ]; diff --git a/master/search/files_4.js b/master/search/files_4.js index 940b9126f4..93ae241fbd 100644 --- a/master/search/files_4.js +++ b/master/search/files_4.js @@ -1,6 +1,6 @@ var searchData= [ - ['eigensolver_2eh_763',['eigensolver.h',['../eigensolver_2eigensolver_8h.html',1,'(Global Namespace)'],['../eigensolver_8h.html',1,'(Global Namespace)']]], - ['enum_5foutput_2eh_764',['enum_output.h',['../blas_2enum__output_8h.html',1,'']]], - ['error_2eh_765',['error.h',['../communication_2error_8h.html',1,'']]] + ['eigensolver_2eh_766',['eigensolver.h',['../eigensolver_2eigensolver_8h.html',1,'(Global Namespace)'],['../eigensolver_8h.html',1,'(Global Namespace)']]], + ['enum_5foutput_2eh_767',['enum_output.h',['../blas_2enum__output_8h.html',1,'']]], + ['error_2eh_768',['error.h',['../communication_2error_8h.html',1,'']]] ]; diff --git a/master/search/files_5.js b/master/search/files_5.js index e574a8dda3..f5e1d19419 100644 --- a/master/search/files_5.js +++ b/master/search/files_5.js @@ -1,6 +1,6 @@ var searchData= [ - ['factorization_2eh_766',['factorization.h',['../factorization_8h.html',1,'']]], - ['format_5fshort_2eh_767',['format_short.h',['../format__short_8h.html',1,'']]], - ['functions_5fsync_2eh_768',['functions_sync.h',['../functions__sync_8h.html',1,'']]] + ['factorization_2eh_769',['factorization.h',['../factorization_8h.html',1,'']]], + ['format_5fshort_2eh_770',['format_short.h',['../format__short_8h.html',1,'']]], + ['functions_5fsync_2eh_771',['functions_sync.h',['../functions__sync_8h.html',1,'']]] ]; diff --git a/master/search/files_6.js b/master/search/files_6.js index f1d7ac052f..107a0a93d3 100644 --- a/master/search/files_6.js +++ b/master/search/files_6.js @@ -1,6 +1,6 @@ var searchData= [ - ['gen_5feigensolver_2eh_769',['gen_eigensolver.h',['../gen__eigensolver_8h.html',1,'']]], - ['gen_5fto_5fstd_2eh_770',['gen_to_std.h',['../gen__to__std_8h.html',1,'']]], - ['general_2eh_771',['general.h',['../multiplication_2general_8h.html',1,'(Global Namespace)'],['../permutations_2general_8h.html',1,'(Global Namespace)']]] + ['gen_5feigensolver_2eh_772',['gen_eigensolver.h',['../gen__eigensolver_8h.html',1,'']]], + ['gen_5fto_5fstd_2eh_773',['gen_to_std.h',['../gen__to__std_8h.html',1,'']]], + ['general_2eh_774',['general.h',['../multiplication_2general_8h.html',1,'(Global Namespace)'],['../permutations_2general_8h.html',1,'(Global Namespace)']]] ]; diff --git a/master/search/files_7.js b/master/search/files_7.js index 8cc7a034a5..653223e42e 100644 --- a/master/search/files_7.js +++ b/master/search/files_7.js @@ -1,4 +1,4 @@ var searchData= [ - ['hermitian_2eh_772',['hermitian.h',['../hermitian_8h.html',1,'']]] + ['hermitian_2eh_775',['hermitian.h',['../hermitian_8h.html',1,'']]] ]; diff --git a/master/search/files_8.js b/master/search/files_8.js index 0c438af520..718042f7c0 100644 --- a/master/search/files_8.js +++ b/master/search/files_8.js @@ -1,6 +1,6 @@ var searchData= [ - ['index_2eh_773',['index.h',['../index_8h.html',1,'']]], - ['index2d_2eh_774',['index2d.h',['../index2d_8h.html',1,'']]], - ['init_2eh_775',['init.h',['../communication_2init_8h.html',1,'(Global Namespace)'],['../init_8h.html',1,'(Global Namespace)']]] + ['index_2eh_776',['index.h',['../index_8h.html',1,'']]], + ['index2d_2eh_777',['index2d.h',['../index2d_8h.html',1,'']]], + ['init_2eh_778',['init.h',['../communication_2init_8h.html',1,'(Global Namespace)'],['../init_8h.html',1,'(Global Namespace)']]] ]; diff --git a/master/search/files_9.js b/master/search/files_9.js index de2bcb8782..fe731f04fe 100644 --- a/master/search/files_9.js +++ b/master/search/files_9.js @@ -1,4 +1,4 @@ var searchData= [ - ['kernels_2eh_776',['kernels.h',['../communication_2kernels_8h.html',1,'']]] + ['kernels_2eh_779',['kernels.h',['../communication_2kernels_8h.html',1,'']]] ]; diff --git a/master/search/files_a.js b/master/search/files_a.js index 0968144c44..502a44dfd5 100644 --- a/master/search/files_a.js +++ b/master/search/files_a.js @@ -1,4 +1,4 @@ var searchData= [ - ['layout_5finfo_2eh_777',['layout_info.h',['../layout__info_8h.html',1,'']]] + ['layout_5finfo_2eh_780',['layout_info.h',['../layout__info_8h.html',1,'']]] ]; diff --git a/master/search/files_b.js b/master/search/files_b.js index 5619eee3ce..ce5f41e9a5 100644 --- a/master/search/files_b.js +++ b/master/search/files_b.js @@ -1,9 +1,9 @@ var searchData= [ - ['matrix_2eh_778',['matrix.h',['../matrix_8h.html',1,'']]], - ['matrix_5fbase_2eh_779',['matrix_base.h',['../matrix__base_8h.html',1,'']]], - ['matrix_5fmirror_2eh_780',['matrix_mirror.h',['../matrix__mirror_8h.html',1,'']]], - ['memory_5fchunk_2eh_781',['memory_chunk.h',['../memory__chunk_8h.html',1,'']]], - ['memory_5fview_2eh_782',['memory_view.h',['../memory__view_8h.html',1,'']]], - ['message_2eh_783',['message.h',['../message_8h.html',1,'']]] + ['matrix_2eh_781',['matrix.h',['../matrix_8h.html',1,'']]], + ['matrix_5fbase_2eh_782',['matrix_base.h',['../matrix__base_8h.html',1,'']]], + ['matrix_5fmirror_2eh_783',['matrix_mirror.h',['../matrix__mirror_8h.html',1,'']]], + ['memory_5fchunk_2eh_784',['memory_chunk.h',['../memory__chunk_8h.html',1,'']]], + ['memory_5fview_2eh_785',['memory_view.h',['../memory__view_8h.html',1,'']]], + ['message_2eh_786',['message.h',['../message_8h.html',1,'']]] ]; diff --git a/master/search/files_c.js b/master/search/files_c.js index 2b58fe0939..05208b2906 100644 --- a/master/search/files_c.js +++ b/master/search/files_c.js @@ -1,4 +1,4 @@ var searchData= [ - ['norm_2eh_784',['norm.h',['../norm_8h.html',1,'']]] + ['norm_2eh_787',['norm.h',['../norm_8h.html',1,'']]] ]; diff --git a/master/search/files_d.js b/master/search/files_d.js index 12701aa695..92897be57b 100644 --- a/master/search/files_d.js +++ b/master/search/files_d.js @@ -1,10 +1,10 @@ var searchData= [ - ['p2p_2eh_785',['p2p.h',['../p2p_8h.html',1,'']]], - ['p2p_5fallsum_2eh_786',['p2p_allsum.h',['../p2p__allsum_8h.html',1,'']]], - ['panel_2eh_787',['panel.h',['../panel_8h.html',1,'']]], - ['pipeline_2eh_788',['pipeline.h',['../pipeline_8h.html',1,'']]], - ['print_5fcsv_2eh_789',['print_csv.h',['../print__csv_8h.html',1,'']]], - ['print_5fgpu_2eh_790',['print_gpu.h',['../print__gpu_8h.html',1,'']]], - ['print_5fnumpy_2eh_791',['print_numpy.h',['../print__numpy_8h.html',1,'']]] + ['p2p_2eh_788',['p2p.h',['../p2p_8h.html',1,'']]], + ['p2p_5fallsum_2eh_789',['p2p_allsum.h',['../p2p__allsum_8h.html',1,'']]], + ['panel_2eh_790',['panel.h',['../panel_8h.html',1,'']]], + ['pipeline_2eh_791',['pipeline.h',['../pipeline_8h.html',1,'']]], + ['print_5fcsv_2eh_792',['print_csv.h',['../print__csv_8h.html',1,'']]], + ['print_5fgpu_2eh_793',['print_gpu.h',['../print__gpu_8h.html',1,'']]], + ['print_5fnumpy_2eh_794',['print_numpy.h',['../print__numpy_8h.html',1,'']]] ]; diff --git a/master/search/files_e.js b/master/search/files_e.js index 61babb9e72..70b06c4b51 100644 --- a/master/search/files_e.js +++ b/master/search/files_e.js @@ -1,4 +1,4 @@ var searchData= [ - ['qr_2eh_792',['qr.h',['../qr_8h.html',1,'']]] + ['qr_2eh_795',['qr.h',['../qr_8h.html',1,'']]] ]; diff --git a/master/search/files_f.js b/master/search/files_f.js index cd579f9cb8..9acc87001d 100644 --- a/master/search/files_f.js +++ b/master/search/files_f.js @@ -1,9 +1,8 @@ var searchData= [ - ['range2d_2eh_793',['range2d.h',['../range2d_8h.html',1,'']]], - ['rdma_2eh_794',['rdma.h',['../rdma_8h.html',1,'']]], - ['reduce_2eh_795',['reduce.h',['../kernels_2reduce_8h.html',1,'(Global Namespace)'],['../sync_2reduce_8h.html',1,'(Global Namespace)']]], - ['reduction_5fto_5fband_2eh_796',['reduction_to_band.h',['../reduction__to__band_8h.html',1,'']]], - ['retiled_5fmatrix_2eh_797',['retiled_matrix.h',['../retiled__matrix_8h.html',1,'']]], - ['round_5frobin_2eh_798',['round_robin.h',['../round__robin_8h.html',1,'']]] + ['range2d_2eh_796',['range2d.h',['../range2d_8h.html',1,'']]], + ['rdma_2eh_797',['rdma.h',['../rdma_8h.html',1,'']]], + ['reduce_2eh_798',['reduce.h',['../kernels_2reduce_8h.html',1,'(Global Namespace)'],['../sync_2reduce_8h.html',1,'(Global Namespace)']]], + ['reduction_5fto_5fband_2eh_799',['reduction_to_band.h',['../reduction__to__band_8h.html',1,'']]], + ['round_5frobin_2eh_800',['round_robin.h',['../round__robin_8h.html',1,'']]] ]; diff --git a/master/search/functions_0.js b/master/search/functions_0.js index bf522edcdb..fffd6efcea 100644 --- a/master/search/functions_0.js +++ b/master/search/functions_0.js @@ -1,6 +1,6 @@ var searchData= [ - ['add_822',['add',['../tile__extensions_8h.html#aaff9e7cefc29db10ec4d5619a723b1e8',1,'dlaf::tile::add(T alpha, const matrix::Tile< const T, D > &tile_b, const matrix::Tile< T, D > &tile_a)'],['../tile__extensions_8h.html#aee89b6690331c2208abce640e9bd81a7',1,'dlaf::tile::add(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../tile__extensions_8h.html#ab6d1e869fabafc8c41981cb76e3bb543',1,'dlaf::tile::add(const dlaf::internal::Policy< B > &p)']]], - ['allreduce_823',['allReduce',['../sync_2all__reduce_8h.html#ab2ce05dba76e64025b65287f83331332',1,'dlaf::comm::sync']]], - ['allreduceinplace_824',['allReduceInPlace',['../sync_2all__reduce_8h.html#ac31a941ab370f1930ce5ed0f0717b00b',1,'dlaf::comm::sync']]] + ['add_824',['add',['../tile__extensions_8h.html#aaff9e7cefc29db10ec4d5619a723b1e8',1,'dlaf::tile::add(T alpha, const matrix::Tile< const T, D > &tile_b, const matrix::Tile< T, D > &tile_a)'],['../tile__extensions_8h.html#aee89b6690331c2208abce640e9bd81a7',1,'dlaf::tile::add(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../tile__extensions_8h.html#ab6d1e869fabafc8c41981cb76e3bb543',1,'dlaf::tile::add(const dlaf::internal::Policy< B > &p)']]], + ['allreduce_825',['allReduce',['../sync_2all__reduce_8h.html#ab2ce05dba76e64025b65287f83331332',1,'dlaf::comm::sync']]], + ['allreduceinplace_826',['allReduceInPlace',['../sync_2all__reduce_8h.html#ac31a941ab370f1930ce5ed0f0717b00b',1,'dlaf::comm::sync']]] ]; diff --git a/master/search/functions_1.js b/master/search/functions_1.js index 77ae9b9d77..2923fc470f 100644 --- a/master/search/functions_1.js +++ b/master/search/functions_1.js @@ -1,11 +1,11 @@ var searchData= [ - ['backtransformationreductiontoband_825',['backTransformationReductionToBand',['../bt__reduction__to__band_8h.html#aa0009f01a9bcae7b94d16e3696433ef8',1,'dlaf::eigensolver::backTransformationReductionToBand(comm::CommunicatorGrid grid, const SizeType b, Matrix< T, device > &mat_c, Matrix< const T, device > &mat_v, Matrix< const T, Device::CPU > &mat_taus)'],['../bt__reduction__to__band_8h.html#a04e50376b3136f312c5c4114b917ae03',1,'dlaf::eigensolver::backTransformationReductionToBand(const SizeType b, Matrix< T, device > &mat_c, Matrix< const T, device > &mat_v, Matrix< const T, Device::CPU > &mat_taus)']]], - ['bandtotridiag_826',['bandToTridiag',['../band__to__tridiag_8h.html#ad81b7972326c06b90ed419d9169351fc',1,'dlaf::eigensolver::bandToTridiag(comm::CommunicatorGrid grid, blas::Uplo uplo, SizeType band_size, Matrix< const T, device > &mat_a)'],['../band__to__tridiag_8h.html#a5389b4333d5882eb71a9ccf6a24b2434',1,'dlaf::eigensolver::bandToTridiag(blas::Uplo uplo, SizeType band_size, Matrix< const T, D > &mat_a)']]], - ['basetilesize_827',['baseTileSize',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#add48022e3fd9298035052a023fa6dbb0',1,'dlaf::matrix::internal::MatrixBase']]], - ['basic_5fcoords_828',['basic_coords',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a80a7c82a6cf0cef1cb765fb62ab42173',1,'dlaf::common::internal::basic_coords::basic_coords(IndexT row, IndexT col) noexcept'],['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a0f13ee129706aa74fefba6cefa29ab55',1,'dlaf::common::internal::basic_coords::basic_coords(Coord component, IndexT value, IndexT fixed=0) noexcept'],['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#ac7a56bcf4ceaad4b19c02b12a9b7e5df',1,'dlaf::common::internal::basic_coords::basic_coords(const std::array< IndexT, 2 > &coords) noexcept'],['../classdlaf_1_1common_1_1_size2_d.html#a80a7c82a6cf0cef1cb765fb62ab42173',1,'dlaf::common::Size2D::basic_coords(IndexT row, IndexT col) noexcept'],['../classdlaf_1_1common_1_1_size2_d.html#a0f13ee129706aa74fefba6cefa29ab55',1,'dlaf::common::Size2D::basic_coords(Coord component, IndexT value, IndexT fixed=0) noexcept'],['../classdlaf_1_1common_1_1_size2_d.html#ac7a56bcf4ceaad4b19c02b12a9b7e5df',1,'dlaf::common::Size2D::basic_coords(const std::array< IndexT, 2 > &coords) noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#a80a7c82a6cf0cef1cb765fb62ab42173',1,'dlaf::common::Index2D::basic_coords(IndexT row, IndexT col) noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#a0f13ee129706aa74fefba6cefa29ab55',1,'dlaf::common::Index2D::basic_coords(Coord component, IndexT value, IndexT fixed=0) noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#ac7a56bcf4ceaad4b19c02b12a9b7e5df',1,'dlaf::common::Index2D::basic_coords(const std::array< IndexT, 2 > &coords) noexcept']]], - ['blocksize_829',['blockSize',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a3f645fd23ce4c20591fdd2738b235f92',1,'dlaf::matrix::internal::MatrixBase']]], - ['blocksize_830',['blocksize',['../structdlaf_1_1common_1_1_data_descriptor.html#aeb70d2395b5e4f49d56e133d3d4b3dd3',1,'dlaf::common::DataDescriptor']]], - ['broadcast_831',['broadcast',['../broadcast__panel_8h.html#a83a59dd8a250067f25b9b4bbd041a9cf',1,'dlaf::comm::broadcast(comm::IndexT_MPI rank_root, matrix::Panel< axis, T, D, storage > &panel, common::Pipeline< comm::Communicator > &serial_comm)'],['../broadcast__panel_8h.html#a6e56082c935ff28b0184870563b4069b',1,'dlaf::comm::broadcast(comm::IndexT_MPI rank_root, matrix::Panel< axis, T, D, storage > &panel, matrix::Panel< orthogonal(axis), T, D, storageT > &panelT, common::Pipeline< comm::Communicator > &row_task_chain, common::Pipeline< comm::Communicator > &col_task_chain)']]], - ['buffer_832',['Buffer',['../structdlaf_1_1common_1_1_buffer.html#a904065ced6826be020d27b87b4d7596d',1,'dlaf::common::Buffer::Buffer(const SizeType N)'],['../structdlaf_1_1common_1_1_buffer.html#a84240f0e2434c412292ed0c1a9db6c61',1,'dlaf::common::Buffer::Buffer(std::unique_ptr< T[]> &&memory, const SizeType N)'],['../structdlaf_1_1common_1_1_buffer.html#ae96ced7d357b20a027c5499aad807293',1,'dlaf::common::Buffer::Buffer()=default']]] + ['backtransformationreductiontoband_827',['backTransformationReductionToBand',['../bt__reduction__to__band_8h.html#aa0009f01a9bcae7b94d16e3696433ef8',1,'dlaf::eigensolver::backTransformationReductionToBand(comm::CommunicatorGrid grid, const SizeType b, Matrix< T, device > &mat_c, Matrix< const T, device > &mat_v, Matrix< const T, Device::CPU > &mat_taus)'],['../bt__reduction__to__band_8h.html#a04e50376b3136f312c5c4114b917ae03',1,'dlaf::eigensolver::backTransformationReductionToBand(const SizeType b, Matrix< T, device > &mat_c, Matrix< const T, device > &mat_v, Matrix< const T, Device::CPU > &mat_taus)']]], + ['bandtotridiag_828',['bandToTridiag',['../band__to__tridiag_8h.html#ad81b7972326c06b90ed419d9169351fc',1,'dlaf::eigensolver::bandToTridiag(comm::CommunicatorGrid grid, blas::Uplo uplo, SizeType band_size, Matrix< const T, device > &mat_a)'],['../band__to__tridiag_8h.html#a5389b4333d5882eb71a9ccf6a24b2434',1,'dlaf::eigensolver::bandToTridiag(blas::Uplo uplo, SizeType band_size, Matrix< const T, D > &mat_a)']]], + ['basetilesize_829',['baseTileSize',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#add48022e3fd9298035052a023fa6dbb0',1,'dlaf::matrix::internal::MatrixBase']]], + ['basic_5fcoords_830',['basic_coords',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a80a7c82a6cf0cef1cb765fb62ab42173',1,'dlaf::common::internal::basic_coords::basic_coords(IndexT row, IndexT col) noexcept'],['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a0f13ee129706aa74fefba6cefa29ab55',1,'dlaf::common::internal::basic_coords::basic_coords(Coord component, IndexT value, IndexT fixed=0) noexcept'],['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#ac7a56bcf4ceaad4b19c02b12a9b7e5df',1,'dlaf::common::internal::basic_coords::basic_coords(const std::array< IndexT, 2 > &coords) noexcept'],['../classdlaf_1_1common_1_1_size2_d.html#a80a7c82a6cf0cef1cb765fb62ab42173',1,'dlaf::common::Size2D::basic_coords(IndexT row, IndexT col) noexcept'],['../classdlaf_1_1common_1_1_size2_d.html#a0f13ee129706aa74fefba6cefa29ab55',1,'dlaf::common::Size2D::basic_coords(Coord component, IndexT value, IndexT fixed=0) noexcept'],['../classdlaf_1_1common_1_1_size2_d.html#ac7a56bcf4ceaad4b19c02b12a9b7e5df',1,'dlaf::common::Size2D::basic_coords(const std::array< IndexT, 2 > &coords) noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#a80a7c82a6cf0cef1cb765fb62ab42173',1,'dlaf::common::Index2D::basic_coords(IndexT row, IndexT col) noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#a0f13ee129706aa74fefba6cefa29ab55',1,'dlaf::common::Index2D::basic_coords(Coord component, IndexT value, IndexT fixed=0) noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#ac7a56bcf4ceaad4b19c02b12a9b7e5df',1,'dlaf::common::Index2D::basic_coords(const std::array< IndexT, 2 > &coords) noexcept']]], + ['blocksize_831',['blockSize',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a3f645fd23ce4c20591fdd2738b235f92',1,'dlaf::matrix::internal::MatrixBase']]], + ['blocksize_832',['blocksize',['../structdlaf_1_1common_1_1_data_descriptor.html#aeb70d2395b5e4f49d56e133d3d4b3dd3',1,'dlaf::common::DataDescriptor']]], + ['broadcast_833',['broadcast',['../broadcast__panel_8h.html#a83a59dd8a250067f25b9b4bbd041a9cf',1,'dlaf::comm::broadcast(comm::IndexT_MPI rank_root, matrix::Panel< axis, T, D, storage > &panel, common::Pipeline< comm::Communicator > &serial_comm)'],['../broadcast__panel_8h.html#a6e56082c935ff28b0184870563b4069b',1,'dlaf::comm::broadcast(comm::IndexT_MPI rank_root, matrix::Panel< axis, T, D, storage > &panel, matrix::Panel< orthogonal(axis), T, D, storageT > &panelT, common::Pipeline< comm::Communicator > &row_task_chain, common::Pipeline< comm::Communicator > &col_task_chain)']]], + ['buffer_834',['Buffer',['../structdlaf_1_1common_1_1_buffer.html#a904065ced6826be020d27b87b4d7596d',1,'dlaf::common::Buffer::Buffer(const SizeType N)'],['../structdlaf_1_1common_1_1_buffer.html#a84240f0e2434c412292ed0c1a9db6c61',1,'dlaf::common::Buffer::Buffer(std::unique_ptr< T[]> &&memory, const SizeType N)'],['../structdlaf_1_1common_1_1_buffer.html#ae96ced7d357b20a027c5499aad807293',1,'dlaf::common::Buffer::Buffer()=default']]] ]; diff --git a/master/search/functions_10.js b/master/search/functions_10.js index c940961679..2c5fded7c4 100644 --- a/master/search/functions_10.js +++ b/master/search/functions_10.js @@ -1,28 +1,28 @@ var searchData= [ - ['tile_1033',['Tile',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a73f2bfb1856011bce752fca6582ba584',1,'dlaf::matrix::Tile< const T, D >::Tile()'],['../classdlaf_1_1matrix_1_1_tile.html#af4bddf624138747a5089071f1f57eee4',1,'dlaf::matrix::Tile::Tile()=default'],['../classdlaf_1_1matrix_1_1_tile.html#a4c7d7a75afcd2cc6911b40643f4f095f',1,'dlaf::matrix::Tile::Tile(const TileElementSize &size, memory::MemoryView< ElementType, D > &&memory_view, SizeType ld) noexcept'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a38b4f570cc1ccedb8880d469d353414e',1,'dlaf::matrix::Tile< const T, D >::Tile()']]], - ['tileelementfromelement_1034',['tileElementFromElement',['../util__distribution_8h.html#a77c1788e03f30e7d9511a7e90a89a494',1,'dlaf::util::matrix']]], - ['tileelementfromglobalelement_1035',['tileElementFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#afe2b66a02c3a1a78bfac1bb2f20f2864',1,'dlaf::matrix::Distribution']]], - ['tileelementindex_1036',['tileElementIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a71ad776a93750b9c6d1ebfa2d36e28db',1,'dlaf::matrix::Distribution']]], - ['tilefromelement_1037',['tileFromElement',['../util__distribution_8h.html#a8876d879740e8f24112ca972961e11c7',1,'dlaf::util::matrix']]], - ['tilelayout_1038',['tileLayout',['../layout__info_8h.html#a873adaaaa2e661caed98af371fbb1943',1,'dlaf::matrix::tileLayout(const matrix::Distribution &distribution)'],['../layout__info_8h.html#a254e1e0948e647642b0f2615aeaa0ea2',1,'dlaf::matrix::tileLayout(const LocalElementSize &size, const TileElementSize &block_size)'],['../layout__info_8h.html#a970afb0f1a7abc87e507e6c4d35f442a',1,'dlaf::matrix::tileLayout(const matrix::Distribution &distribution, SizeType ld_tile, SizeType tiles_per_col)'],['../layout__info_8h.html#a94ba1a30c8eef4798f864ae119c77113',1,'dlaf::matrix::tileLayout(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col)']]], - ['tilelinearindex_1039',['tileLinearIndex',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a5cdaa0376724e6f624591a6a6bcf46d9',1,'dlaf::matrix::internal::MatrixBase']]], - ['tileoffset_1040',['tileOffset',['../classdlaf_1_1matrix_1_1_layout_info.html#ad03e1291efd2ed416ace133a24a30ac4',1,'dlaf::matrix::LayoutInfo']]], - ['tilesize_1041',['tileSize',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a891c1da69cefcfdfb73b3915ccc2cce5',1,'dlaf::matrix::internal::MatrixBase::tileSize()'],['../classdlaf_1_1matrix_1_1_layout_info.html#a62e3619b9016e9c827c120fce629d08e',1,'dlaf::matrix::LayoutInfo::tileSize()'],['../classdlaf_1_1matrix_1_1_distribution.html#a4cec137b1316902f68b63148d282be62',1,'dlaf::matrix::Distribution::tileSize(const GlobalTileIndex &index) const noexcept']]], - ['tilesizefromglobalelement_1042',['tileSizeFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a3dd77d459960a11b56fc1e7fb27932f7',1,'dlaf::matrix::Distribution']]], - ['to_5fint_1043',['to_int',['../types_8h.html#af90e7fd5acadf5987b7199b0bd44deea',1,'dlaf']]], - ['to_5fsigned_1044',['to_signed',['../types_8h.html#aa39d2d6e54b007c18e1c57ff380c73cd',1,'dlaf::to_signed(const U unsigned_value)'],['../types_8h.html#a2246feaf3c49f3f3ae5928082363f7d3',1,'dlaf::to_signed(const SB value)']]], - ['to_5fsizet_1045',['to_sizet',['../types_8h.html#a25d16ff6621f011d9a98be3d9ebf29f1',1,'dlaf']]], - ['to_5fsizetype_1046',['to_SizeType',['../types_8h.html#ab1f0edc8e1281293c9c1325cb9d4bec9',1,'dlaf']]], - ['to_5fuint_1047',['to_uint',['../types_8h.html#a4543febf91f08c23fd8a6eeca8f49d82',1,'dlaf']]], - ['to_5funsigned_1048',['to_unsigned',['../types_8h.html#afeb5db746331f9ca3a669373e1af84ba',1,'dlaf::to_unsigned(const S signed_value)'],['../types_8h.html#ae9973646cf4ce493ed1831a5bdbf7500',1,'dlaf::to_unsigned(const UB unsigned_value)']]], - ['total_5fops_1049',['total_ops',['../types_8h.html#a831d9cfa2a32641121185df5d9d70cdc',1,'dlaf']]], - ['transpose_1050',['transpose',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a98f8b3b0cd7b4c1fa62b40b9de0041bb',1,'dlaf::common::internal::basic_coords']]], - ['transposed_1051',['transposed',['../index2d_8h.html#a024753ac73fbefa174548c575327862b',1,'dlaf::common']]], - ['triangular_1052',['triangular',['../multiplication_2triangular_8h.html#a6300b8eb1a939512a5b1c45286e89f51',1,'dlaf::multiplication::triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../multiplication_2triangular_8h.html#ad4de985e76a6c12c487b36fa773b870b',1,'dlaf::multiplication::triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../solver_2triangular_8h.html#a9d48ec5978acfd3714a66613678a1afd',1,'dlaf::solver::triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../solver_2triangular_8h.html#a3ff8022461abe72c4197e34601d592a9',1,'dlaf::solver::triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)']]], - ['tridiagsolver_1053',['tridiagSolver',['../tridiag__solver_8h.html#a52f0bd4b681a2d1c3e47bfd6a6ebf15a',1,'dlaf::eigensolver::tridiagSolver(Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, device > &evals, Matrix< T, device > &evecs)'],['../tridiag__solver_8h.html#a0cfeab14e9e1dc7a2fb703006cd7e957',1,'dlaf::eigensolver::tridiagSolver(comm::CommunicatorGrid grid, Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, D > &evals, Matrix< T, D > &evecs)']]], - ['trmm_1054',['trmm',['../blas_2tile_8h.html#a383ebe00fbfd1f328dd0378b8aa42b89',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#a185aa2ac627de6e9e393240cda2b5f4c',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a61397e6748a24f13ee6ad7f23e5339f7',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< T, D > &b)']]], - ['trmm3_1055',['trmm3',['../blas_2tile_8h.html#a0cdf7b22f8a310978252ff7a0c456b98',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a9906093b08eefa0a5341b140ce6100a3',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#aebff48d52916dafd14479dc3b1e04e8e',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const Tile< T, D > &c)']]], - ['trsm_1056',['trsm',['../blas_2tile_8h.html#a1b11d671c920bf9dce76092e6d1aaa8c',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a23eb607d942637d4e14331a2f0f253f7',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#a2fe48f2d3a3a7515a99805d135c5b2ce',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< T, D > &b)']]], - ['type_5fhandler_1057',['type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#aa71d40004e1c645161b1d00b511ad9ee',1,'dlaf::comm::internal::type_handler::type_handler() noexcept=default'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#aa9eebfb39b8aabe2344f61a7bf26ee40',1,'dlaf::comm::internal::type_handler::type_handler(SizeType nblocks, SizeType block_size, SizeType stride)']]] + ['tile_1039',['Tile',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a73f2bfb1856011bce752fca6582ba584',1,'dlaf::matrix::Tile< const T, D >::Tile()'],['../classdlaf_1_1matrix_1_1_tile.html#af4bddf624138747a5089071f1f57eee4',1,'dlaf::matrix::Tile::Tile()=default'],['../classdlaf_1_1matrix_1_1_tile.html#a4c7d7a75afcd2cc6911b40643f4f095f',1,'dlaf::matrix::Tile::Tile(const TileElementSize &size, memory::MemoryView< ElementType, D > &&memory_view, SizeType ld) noexcept'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a38b4f570cc1ccedb8880d469d353414e',1,'dlaf::matrix::Tile< const T, D >::Tile()']]], + ['tileelementfromelement_1040',['tileElementFromElement',['../util__distribution_8h.html#a77c1788e03f30e7d9511a7e90a89a494',1,'dlaf::util::matrix']]], + ['tileelementfromglobalelement_1041',['tileElementFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#afe2b66a02c3a1a78bfac1bb2f20f2864',1,'dlaf::matrix::Distribution']]], + ['tileelementindex_1042',['tileElementIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a71ad776a93750b9c6d1ebfa2d36e28db',1,'dlaf::matrix::Distribution']]], + ['tilefromelement_1043',['tileFromElement',['../util__distribution_8h.html#a8876d879740e8f24112ca972961e11c7',1,'dlaf::util::matrix']]], + ['tilelayout_1044',['tileLayout',['../layout__info_8h.html#a873adaaaa2e661caed98af371fbb1943',1,'dlaf::matrix::tileLayout(const matrix::Distribution &distribution)'],['../layout__info_8h.html#a254e1e0948e647642b0f2615aeaa0ea2',1,'dlaf::matrix::tileLayout(const LocalElementSize &size, const TileElementSize &block_size)'],['../layout__info_8h.html#a970afb0f1a7abc87e507e6c4d35f442a',1,'dlaf::matrix::tileLayout(const matrix::Distribution &distribution, SizeType ld_tile, SizeType tiles_per_col)'],['../layout__info_8h.html#a94ba1a30c8eef4798f864ae119c77113',1,'dlaf::matrix::tileLayout(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col)']]], + ['tilelinearindex_1045',['tileLinearIndex',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a5cdaa0376724e6f624591a6a6bcf46d9',1,'dlaf::matrix::internal::MatrixBase']]], + ['tileoffset_1046',['tileOffset',['../classdlaf_1_1matrix_1_1_layout_info.html#ad03e1291efd2ed416ace133a24a30ac4',1,'dlaf::matrix::LayoutInfo']]], + ['tilesize_1047',['tileSize',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a891c1da69cefcfdfb73b3915ccc2cce5',1,'dlaf::matrix::internal::MatrixBase::tileSize()'],['../classdlaf_1_1matrix_1_1_layout_info.html#a62e3619b9016e9c827c120fce629d08e',1,'dlaf::matrix::LayoutInfo::tileSize()'],['../classdlaf_1_1matrix_1_1_distribution.html#a4cec137b1316902f68b63148d282be62',1,'dlaf::matrix::Distribution::tileSize(const GlobalTileIndex &index) const noexcept']]], + ['tilesizefromglobalelement_1048',['tileSizeFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a3dd77d459960a11b56fc1e7fb27932f7',1,'dlaf::matrix::Distribution']]], + ['to_5fint_1049',['to_int',['../types_8h.html#af90e7fd5acadf5987b7199b0bd44deea',1,'dlaf']]], + ['to_5fsigned_1050',['to_signed',['../types_8h.html#aa39d2d6e54b007c18e1c57ff380c73cd',1,'dlaf::to_signed(const U unsigned_value)'],['../types_8h.html#a2246feaf3c49f3f3ae5928082363f7d3',1,'dlaf::to_signed(const SB value)']]], + ['to_5fsizet_1051',['to_sizet',['../types_8h.html#a25d16ff6621f011d9a98be3d9ebf29f1',1,'dlaf']]], + ['to_5fsizetype_1052',['to_SizeType',['../types_8h.html#ab1f0edc8e1281293c9c1325cb9d4bec9',1,'dlaf']]], + ['to_5fuint_1053',['to_uint',['../types_8h.html#a4543febf91f08c23fd8a6eeca8f49d82',1,'dlaf']]], + ['to_5funsigned_1054',['to_unsigned',['../types_8h.html#afeb5db746331f9ca3a669373e1af84ba',1,'dlaf::to_unsigned(const S signed_value)'],['../types_8h.html#ae9973646cf4ce493ed1831a5bdbf7500',1,'dlaf::to_unsigned(const UB unsigned_value)']]], + ['total_5fops_1055',['total_ops',['../types_8h.html#a831d9cfa2a32641121185df5d9d70cdc',1,'dlaf']]], + ['transpose_1056',['transpose',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a98f8b3b0cd7b4c1fa62b40b9de0041bb',1,'dlaf::common::internal::basic_coords']]], + ['transposed_1057',['transposed',['../index2d_8h.html#a024753ac73fbefa174548c575327862b',1,'dlaf::common']]], + ['triangular_1058',['triangular',['../multiplication_2triangular_8h.html#a6300b8eb1a939512a5b1c45286e89f51',1,'dlaf::multiplication::triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../multiplication_2triangular_8h.html#ad4de985e76a6c12c487b36fa773b870b',1,'dlaf::multiplication::triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../solver_2triangular_8h.html#a9d48ec5978acfd3714a66613678a1afd',1,'dlaf::solver::triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)'],['../solver_2triangular_8h.html#a3ff8022461abe72c4197e34601d592a9',1,'dlaf::solver::triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha, Matrix< const T, device > &mat_a, Matrix< T, device > &mat_b)']]], + ['tridiagsolver_1059',['tridiagSolver',['../tridiag__solver_8h.html#a52f0bd4b681a2d1c3e47bfd6a6ebf15a',1,'dlaf::eigensolver::tridiagSolver(Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, device > &evals, Matrix< T, device > &evecs)'],['../tridiag__solver_8h.html#a0cfeab14e9e1dc7a2fb703006cd7e957',1,'dlaf::eigensolver::tridiagSolver(comm::CommunicatorGrid grid, Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, D > &evals, Matrix< T, D > &evecs)']]], + ['trmm_1060',['trmm',['../blas_2tile_8h.html#a383ebe00fbfd1f328dd0378b8aa42b89',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#a185aa2ac627de6e9e393240cda2b5f4c',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a61397e6748a24f13ee6ad7f23e5339f7',1,'dlaf::tile::trmm(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< T, D > &b)']]], + ['trmm3_1061',['trmm3',['../blas_2tile_8h.html#a0cdf7b22f8a310978252ff7a0c456b98',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a9906093b08eefa0a5341b140ce6100a3',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#aebff48d52916dafd14479dc3b1e04e8e',1,'dlaf::tile::trmm3(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const Tile< T, D > &c)']]], + ['trsm_1062',['trsm',['../blas_2tile_8h.html#a1b11d671c920bf9dce76092e6d1aaa8c',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a23eb607d942637d4e14331a2f0f253f7',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#a2fe48f2d3a3a7515a99805d135c5b2ce',1,'dlaf::tile::trsm(const dlaf::internal::Policy< B > &policy, const blas::Side side, const blas::Uplo uplo, const blas::Op op, const blas::Diag diag, const T alpha, const Tile< const T, D > &a, const Tile< T, D > &b)']]], + ['type_5fhandler_1063',['type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#aa71d40004e1c645161b1d00b511ad9ee',1,'dlaf::comm::internal::type_handler::type_handler() noexcept=default'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#aa9eebfb39b8aabe2344f61a7bf26ee40',1,'dlaf::comm::internal::type_handler::type_handler(SizeType nblocks, SizeType block_size, SizeType stride)']]] ]; diff --git a/master/search/functions_11.js b/master/search/functions_11.js index 0d08b82294..bc55d253c5 100644 --- a/master/search/functions_11.js +++ b/master/search/functions_11.js @@ -1,4 +1,4 @@ var searchData= [ - ['unwrap_1058',['unwrap',['../unwrap_8h.html#a8da70a5535ac987a8f6ca7f8b195547d',1,'dlaf::common::internal']]] + ['unwrap_1064',['unwrap',['../unwrap_8h.html#a8da70a5535ac987a8f6ca7f8b195547d',1,'dlaf::common::internal']]] ]; diff --git a/master/search/functions_12.js b/master/search/functions_12.js index 4bd74569a7..9bb3d270a1 100644 --- a/master/search/functions_12.js +++ b/master/search/functions_12.js @@ -1,5 +1,5 @@ var searchData= [ - ['valid_1059',['valid',['../classdlaf_1_1common_1_1_pipeline.html#affda528ea4c5e0e54f8fca4250b53af5',1,'dlaf::common::Pipeline::valid()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a3e3096bb1ef85b6017bd8c42c9617586',1,'dlaf::matrix::internal::TilePipeline::valid()']]], - ['view_1060',['View',['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a3635af3164815f905a9fd33979a14a34',1,'dlaf::matrix::internal::View']]] + ['valid_1065',['valid',['../classdlaf_1_1common_1_1_pipeline.html#affda528ea4c5e0e54f8fca4250b53af5',1,'dlaf::common::Pipeline::valid()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a3e3096bb1ef85b6017bd8c42c9617586',1,'dlaf::matrix::internal::TilePipeline::valid()']]], + ['view_1066',['View',['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a3635af3164815f905a9fd33979a14a34',1,'dlaf::matrix::internal::View']]] ]; diff --git a/master/search/functions_13.js b/master/search/functions_13.js index 9ea6c3b8bd..41f8134b38 100644 --- a/master/search/functions_13.js +++ b/master/search/functions_13.js @@ -1,4 +1,4 @@ var searchData= [ - ['waitlocaltiles_1061',['waitLocalTiles',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#abffc73bf079f7f7c23af3fca4dec329f',1,'dlaf::matrix::Matrix< const T, D >::waitLocalTiles()'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#ab731cfea1c5b07288bc3b34c59cb37d0',1,'dlaf::matrix::RetiledMatrix::waitLocalTiles()']]] + ['waitlocaltiles_1067',['waitLocalTiles',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#abffc73bf079f7f7c23af3fca4dec329f',1,'dlaf::matrix::Matrix< const T, D >']]] ]; diff --git a/master/search/functions_14.js b/master/search/functions_14.js index 9ee18a9282..0438f1acd1 100644 --- a/master/search/functions_14.js +++ b/master/search/functions_14.js @@ -1,9 +1,9 @@ var searchData= [ - ['_7ematrixmirror_1062',['~MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#a2eeef61985e20a9218d55e76548acfb5',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::~MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#ae24915137508619315773c7a416f902f',1,'dlaf::matrix::MatrixMirror::~MatrixMirror()']]], - ['_7ememorychunk_1063',['~MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html#a597aa1c075db44808f7e14c9ec03e50b',1,'dlaf::memory::MemoryChunk']]], - ['_7empi_5finit_1064',['~mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html#a2413c9dc23983c5790cd0ebb3bb9e057',1,'dlaf::comm::mpi_init']]], - ['_7epanel_1065',['~Panel',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#adab53d11d9904715b775df7c23ba6682',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['_7etile_1066',['~Tile',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a66f6b8ef0f4d9168c12715d0fc1b16dc',1,'dlaf::matrix::Tile< const T, D >']]], - ['_7etype_5fhandler_1067',['~type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a8fab81a168aa4f3441deb1a01f4a5171',1,'dlaf::comm::internal::type_handler']]] + ['_7ematrixmirror_1068',['~MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#a2eeef61985e20a9218d55e76548acfb5',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::~MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#ae24915137508619315773c7a416f902f',1,'dlaf::matrix::MatrixMirror::~MatrixMirror()']]], + ['_7ememorychunk_1069',['~MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html#a597aa1c075db44808f7e14c9ec03e50b',1,'dlaf::memory::MemoryChunk']]], + ['_7empi_5finit_1070',['~mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html#a2413c9dc23983c5790cd0ebb3bb9e057',1,'dlaf::comm::mpi_init']]], + ['_7epanel_1071',['~Panel',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#adab53d11d9904715b775df7c23ba6682',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['_7etile_1072',['~Tile',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a66f6b8ef0f4d9168c12715d0fc1b16dc',1,'dlaf::matrix::Tile< const T, D >']]], + ['_7etype_5fhandler_1073',['~type_handler',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a8fab81a168aa4f3441deb1a01f4a5171',1,'dlaf::comm::internal::type_handler']]] ]; diff --git a/master/search/functions_2.js b/master/search/functions_2.js index ce995d246f..1806efd95f 100644 --- a/master/search/functions_2.js +++ b/master/search/functions_2.js @@ -1,29 +1,29 @@ var searchData= [ - ['call_833',['call',['../structdlaf_1_1factorization_1_1internal_1_1_q_r___tfactor.html#ac73f315af163e45f509e230f362d1310',1,'dlaf::factorization::internal::QR_Tfactor::call(matrix::Panel< Coord::Col, T, device > &hh_panel, matrix::ReadOnlyTileSender< T, Device::CPU > taus, matrix::ReadWriteTileSender< T, device > t, common::Pipeline< comm::Communicator > &mpi_col_task_chain)'],['../structdlaf_1_1factorization_1_1internal_1_1_q_r___tfactor.html#a1d9391570e947628f8c23668f944f7d4',1,'dlaf::factorization::internal::QR_Tfactor::call(matrix::Panel< Coord::Col, T, device > &panel_view, matrix::ReadOnlyTileSender< T, Device::CPU > taus, matrix::ReadWriteTileSender< T, device > t)']]], - ['ceildiv_834',['ceilDiv',['../util__math_8h.html#a65aa2478fed7af8bdb404930d7a90fb3',1,'dlaf::util']]], - ['cholesky_835',['cholesky',['../cholesky_8h.html#aed7e035a7eff7eb31eea293621b0c8fa',1,'dlaf::factorization::cholesky(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, device > &mat_a)'],['../cholesky_8h.html#a016f168408d2fe4c2803538e486d7241',1,'dlaf::factorization::cholesky(blas::Uplo uplo, Matrix< T, device > &mat_a)']]], - ['clone_836',['clone',['../classdlaf_1_1comm_1_1_communicator.html#a436eca0d1c23635a6dcd209f1a1c37df',1,'dlaf::comm::Communicator']]], - ['colcommunicator_837',['colCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#af6821b23aeb96284b9e217a02c5115e5',1,'dlaf::comm::CommunicatorGrid']]], - ['colmajorlayout_838',['colMajorLayout',['../layout__info_8h.html#ae46d78a3ef61da55b5c5d70329cc0d11',1,'dlaf::matrix::colMajorLayout(const matrix::Distribution &distribution, SizeType ld)'],['../layout__info_8h.html#adf298ab87e69a013e5f06856ab7e7902',1,'dlaf::matrix::colMajorLayout(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld)']]], - ['cols_839',['cols',['../structdlaf_1_1matrix_1_1_sub_panel_view.html#aa36196de77e52108028c96e7d94e988d',1,'dlaf::matrix::SubPanelView']]], - ['commgridsize_840',['commGridSize',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a1f81375953df07b80ddb5cae672319bd',1,'dlaf::matrix::internal::MatrixBase']]], - ['communicator_841',['Communicator',['../classdlaf_1_1comm_1_1_communicator.html#a891a7bf00a52071373bfc38aae8f83aa',1,'dlaf::comm::Communicator::Communicator()'],['../classdlaf_1_1comm_1_1_communicator.html#afa39beca08a343e1b8587b98d63870a0',1,'dlaf::comm::Communicator::Communicator(MPI_Comm mpi_communicator)']]], - ['communicatorgrid_842',['CommunicatorGrid',['../classdlaf_1_1comm_1_1_communicator_grid.html#aa17d9edca4165db00ede2fa650463aca',1,'dlaf::comm::CommunicatorGrid::CommunicatorGrid(Communicator comm, IndexT_MPI rows, IndexT_MPI cols, common::Ordering ordering)'],['../classdlaf_1_1comm_1_1_communicator_grid.html#ab8d976dcbcf87aa12cde5433fdcbe8a7',1,'dlaf::comm::CommunicatorGrid::CommunicatorGrid(Communicator comm, const std::array< IndexT_MPI, 2 > &size, common::Ordering ordering)']]], - ['computecoords_843',['computeCoords',['../index2d_8h.html#ab6f5ed39b526ad9ae94ed6394a35832b',1,'dlaf::common']]], - ['computecoordscolmajor_844',['computeCoordsColMajor',['../index2d_8h.html#afa922ad7c9bdb475884943a4cb233744',1,'dlaf::common']]], - ['computecoordsrowmajor_845',['computeCoordsRowMajor',['../index2d_8h.html#ac9c99027fa12a875fa12d878393fefff',1,'dlaf::common']]], - ['computelinearindex_846',['computeLinearIndex',['../index2d_8h.html#a15f2674deab8ea5e84471b18b8ffba94',1,'dlaf::common']]], - ['computelinearindexcolmajor_847',['computeLinearIndexColMajor',['../index2d_8h.html#a7e8f91611ddcaf36316ee316666e4d07',1,'dlaf::common']]], - ['computelinearindexrowmajor_848',['computeLinearIndexRowMajor',['../index2d_8h.html#ad3fcc947c550d2315d1afca1241b29b9',1,'dlaf::common']]], - ['computetfactor_849',['computeTFactor',['../qr_8h.html#a4654c00202b44c6c9365d1a03eff9e99',1,'dlaf::factorization::internal']]], - ['conj_850',['conj',['../types_8h.html#a669723b711f94043da9793d2b24d51d9',1,'dlaf::conj(const std::complex< T > number)'],['../types_8h.html#aa28e454906f2ff49398f76a2be8789a4',1,'dlaf::conj(const T number)']]], - ['copy_851',['copy',['../copy__tile_8h.html#a6b3c08cc454ab1e1f36e53b59c667c2f',1,'dlaf::matrix::internal::copy()'],['../data_8h.html#a7f6919f56260399a905af9e869afba10',1,'dlaf::common::copy()'],['../copy_8h.html#abc90df94f398c001b29a9d62a5714043',1,'dlaf::matrix::copy()']]], - ['copysourcetotarget_852',['copySourceToTarget',['../classdlaf_1_1matrix_1_1_matrix_mirror.html#af6a0069ca18c893cd63044f09df006a8',1,'dlaf::matrix::MatrixMirror::copySourceToTarget()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a928052751a50582ba6d7f793fdc877d3',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::copySourceToTarget()']]], - ['copytargettosource_853',['copyTargetToSource',['../classdlaf_1_1matrix_1_1_matrix_mirror.html#abb91b94a9c46d72c9e0f2566bf34728c',1,'dlaf::matrix::MatrixMirror::copyTargetToSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a05e120862eb3eae604143d67afef22bf',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::copyTargetToSource()']]], - ['count_854',['count',['../structdlaf_1_1common_1_1_data_descriptor.html#a936d7432ec4732975764786e9e74cf39',1,'dlaf::common::DataDescriptor::count()'],['../classdlaf_1_1comm_1_1_message.html#a53b9c220b5c7a92bee1a5ad0963014d4',1,'dlaf::comm::Message::count()']]], - ['create_5fdata_855',['create_data',['../data_8h.html#ab547927373241ec4d208e6e62b1b7f80',1,'dlaf::common::create_data()'],['../matrix_2tile_8h.html#a5635e7d6d98ad8b7bcfacfc502e6bdee',1,'dlaf::matrix::create_data()']]], - ['create_5ftemporary_5fbuffer_856',['create_temporary_buffer',['../data__descriptor_8h.html#a7168d5e011929eeafdfebaee7492c9ba',1,'dlaf::common']]], - ['creatematrixfromcolmajor_857',['createMatrixFromColMajor',['../matrix_8h.html#a0edf764f5d64e80355ffe9a5c8eb31e7',1,'dlaf::matrix::createMatrixFromColMajor(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld, T *ptr)'],['../matrix_8h.html#aace8acce04f452c07b86ea6d04b33366',1,'dlaf::matrix::createMatrixFromColMajor(const GlobalElementSize &size, const TileElementSize &block_size, SizeType ld, const comm::CommunicatorGrid &comm, const comm::Index2D &source_rank_index, T *ptr)'],['../matrix_8h.html#a6191cb29721235980256d98eeea8d722',1,'dlaf::matrix::createMatrixFromColMajor(const GlobalElementSize &size, const TileElementSize &block_size, SizeType ld, const comm::CommunicatorGrid &comm, T *ptr)']]], - ['creatematrixfromtile_858',['createMatrixFromTile',['../matrix_8h.html#abae8f550140d87bbe74dfb9bc35c5226',1,'dlaf::matrix::createMatrixFromTile(const LocalElementSize &size, const TileElementSize &block_size, T *ptr)'],['../matrix_8h.html#a5b237a24c0c9c1d9ca958e56684973ae',1,'dlaf::matrix::createMatrixFromTile(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col, T *ptr)'],['../matrix_8h.html#a56931003af5239d82bb1ba403af878e4',1,'dlaf::matrix::createMatrixFromTile(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm, const comm::Index2D &source_rank_index, T *ptr)'],['../matrix_8h.html#a26c92f6737011495d6b3764725c4949a',1,'dlaf::matrix::createMatrixFromTile(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm, T *ptr)'],['../matrix_8h.html#a6bbd903412fe86fb4e7ad4b9b52a2102',1,'dlaf::matrix::createMatrixFromTile(const GlobalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col, const comm::CommunicatorGrid &comm, const comm::Index2D &source_rank_index, T *ptr)'],['../matrix_8h.html#a71205e9a339140e0f9844c929d963d85',1,'dlaf::matrix::createMatrixFromTile(const GlobalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col, const comm::CommunicatorGrid &comm, T *ptr)']]] + ['call_835',['call',['../structdlaf_1_1factorization_1_1internal_1_1_q_r___tfactor.html#ac73f315af163e45f509e230f362d1310',1,'dlaf::factorization::internal::QR_Tfactor::call(matrix::Panel< Coord::Col, T, device > &hh_panel, matrix::ReadOnlyTileSender< T, Device::CPU > taus, matrix::ReadWriteTileSender< T, device > t, common::Pipeline< comm::Communicator > &mpi_col_task_chain)'],['../structdlaf_1_1factorization_1_1internal_1_1_q_r___tfactor.html#a1d9391570e947628f8c23668f944f7d4',1,'dlaf::factorization::internal::QR_Tfactor::call(matrix::Panel< Coord::Col, T, device > &panel_view, matrix::ReadOnlyTileSender< T, Device::CPU > taus, matrix::ReadWriteTileSender< T, device > t)']]], + ['ceildiv_836',['ceilDiv',['../util__math_8h.html#a65aa2478fed7af8bdb404930d7a90fb3',1,'dlaf::util']]], + ['cholesky_837',['cholesky',['../cholesky_8h.html#aed7e035a7eff7eb31eea293621b0c8fa',1,'dlaf::factorization::cholesky(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, device > &mat_a)'],['../cholesky_8h.html#a016f168408d2fe4c2803538e486d7241',1,'dlaf::factorization::cholesky(blas::Uplo uplo, Matrix< T, device > &mat_a)']]], + ['clone_838',['clone',['../classdlaf_1_1comm_1_1_communicator.html#a436eca0d1c23635a6dcd209f1a1c37df',1,'dlaf::comm::Communicator']]], + ['colcommunicator_839',['colCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#af6821b23aeb96284b9e217a02c5115e5',1,'dlaf::comm::CommunicatorGrid']]], + ['colmajorlayout_840',['colMajorLayout',['../layout__info_8h.html#ae46d78a3ef61da55b5c5d70329cc0d11',1,'dlaf::matrix::colMajorLayout(const matrix::Distribution &distribution, SizeType ld)'],['../layout__info_8h.html#adf298ab87e69a013e5f06856ab7e7902',1,'dlaf::matrix::colMajorLayout(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld)']]], + ['cols_841',['cols',['../structdlaf_1_1matrix_1_1_sub_panel_view.html#aa36196de77e52108028c96e7d94e988d',1,'dlaf::matrix::SubPanelView']]], + ['commgridsize_842',['commGridSize',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a1f81375953df07b80ddb5cae672319bd',1,'dlaf::matrix::internal::MatrixBase']]], + ['communicator_843',['Communicator',['../classdlaf_1_1comm_1_1_communicator.html#a891a7bf00a52071373bfc38aae8f83aa',1,'dlaf::comm::Communicator::Communicator()'],['../classdlaf_1_1comm_1_1_communicator.html#afa39beca08a343e1b8587b98d63870a0',1,'dlaf::comm::Communicator::Communicator(MPI_Comm mpi_communicator)']]], + ['communicatorgrid_844',['CommunicatorGrid',['../classdlaf_1_1comm_1_1_communicator_grid.html#aa17d9edca4165db00ede2fa650463aca',1,'dlaf::comm::CommunicatorGrid::CommunicatorGrid(Communicator comm, IndexT_MPI rows, IndexT_MPI cols, common::Ordering ordering)'],['../classdlaf_1_1comm_1_1_communicator_grid.html#ab8d976dcbcf87aa12cde5433fdcbe8a7',1,'dlaf::comm::CommunicatorGrid::CommunicatorGrid(Communicator comm, const std::array< IndexT_MPI, 2 > &size, common::Ordering ordering)']]], + ['computecoords_845',['computeCoords',['../index2d_8h.html#ab6f5ed39b526ad9ae94ed6394a35832b',1,'dlaf::common']]], + ['computecoordscolmajor_846',['computeCoordsColMajor',['../index2d_8h.html#afa922ad7c9bdb475884943a4cb233744',1,'dlaf::common']]], + ['computecoordsrowmajor_847',['computeCoordsRowMajor',['../index2d_8h.html#ac9c99027fa12a875fa12d878393fefff',1,'dlaf::common']]], + ['computelinearindex_848',['computeLinearIndex',['../index2d_8h.html#a15f2674deab8ea5e84471b18b8ffba94',1,'dlaf::common']]], + ['computelinearindexcolmajor_849',['computeLinearIndexColMajor',['../index2d_8h.html#a7e8f91611ddcaf36316ee316666e4d07',1,'dlaf::common']]], + ['computelinearindexrowmajor_850',['computeLinearIndexRowMajor',['../index2d_8h.html#ad3fcc947c550d2315d1afca1241b29b9',1,'dlaf::common']]], + ['computetfactor_851',['computeTFactor',['../qr_8h.html#a4654c00202b44c6c9365d1a03eff9e99',1,'dlaf::factorization::internal']]], + ['conj_852',['conj',['../types_8h.html#a669723b711f94043da9793d2b24d51d9',1,'dlaf::conj(const std::complex< T > number)'],['../types_8h.html#aa28e454906f2ff49398f76a2be8789a4',1,'dlaf::conj(const T number)']]], + ['copy_853',['copy',['../copy__tile_8h.html#a6b3c08cc454ab1e1f36e53b59c667c2f',1,'dlaf::matrix::internal::copy()'],['../data_8h.html#a7f6919f56260399a905af9e869afba10',1,'dlaf::common::copy()'],['../copy_8h.html#abc90df94f398c001b29a9d62a5714043',1,'dlaf::matrix::copy()']]], + ['copysourcetotarget_854',['copySourceToTarget',['../classdlaf_1_1matrix_1_1_matrix_mirror.html#af6a0069ca18c893cd63044f09df006a8',1,'dlaf::matrix::MatrixMirror::copySourceToTarget()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a928052751a50582ba6d7f793fdc877d3',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::copySourceToTarget()']]], + ['copytargettosource_855',['copyTargetToSource',['../classdlaf_1_1matrix_1_1_matrix_mirror.html#abb91b94a9c46d72c9e0f2566bf34728c',1,'dlaf::matrix::MatrixMirror::copyTargetToSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a05e120862eb3eae604143d67afef22bf',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::copyTargetToSource()']]], + ['count_856',['count',['../structdlaf_1_1common_1_1_data_descriptor.html#a936d7432ec4732975764786e9e74cf39',1,'dlaf::common::DataDescriptor::count()'],['../classdlaf_1_1comm_1_1_message.html#a53b9c220b5c7a92bee1a5ad0963014d4',1,'dlaf::comm::Message::count()']]], + ['create_5fdata_857',['create_data',['../data_8h.html#ab547927373241ec4d208e6e62b1b7f80',1,'dlaf::common::create_data()'],['../matrix_2tile_8h.html#a5635e7d6d98ad8b7bcfacfc502e6bdee',1,'dlaf::matrix::create_data()']]], + ['create_5ftemporary_5fbuffer_858',['create_temporary_buffer',['../data__descriptor_8h.html#a7168d5e011929eeafdfebaee7492c9ba',1,'dlaf::common']]], + ['creatematrixfromcolmajor_859',['createMatrixFromColMajor',['../matrix_8h.html#a0edf764f5d64e80355ffe9a5c8eb31e7',1,'dlaf::matrix::createMatrixFromColMajor(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld, T *ptr)'],['../matrix_8h.html#aace8acce04f452c07b86ea6d04b33366',1,'dlaf::matrix::createMatrixFromColMajor(const GlobalElementSize &size, const TileElementSize &block_size, SizeType ld, const comm::CommunicatorGrid &comm, const comm::Index2D &source_rank_index, T *ptr)'],['../matrix_8h.html#a6191cb29721235980256d98eeea8d722',1,'dlaf::matrix::createMatrixFromColMajor(const GlobalElementSize &size, const TileElementSize &block_size, SizeType ld, const comm::CommunicatorGrid &comm, T *ptr)']]], + ['creatematrixfromtile_860',['createMatrixFromTile',['../matrix_8h.html#abae8f550140d87bbe74dfb9bc35c5226',1,'dlaf::matrix::createMatrixFromTile(const LocalElementSize &size, const TileElementSize &block_size, T *ptr)'],['../matrix_8h.html#a5b237a24c0c9c1d9ca958e56684973ae',1,'dlaf::matrix::createMatrixFromTile(const LocalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col, T *ptr)'],['../matrix_8h.html#a56931003af5239d82bb1ba403af878e4',1,'dlaf::matrix::createMatrixFromTile(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm, const comm::Index2D &source_rank_index, T *ptr)'],['../matrix_8h.html#a26c92f6737011495d6b3764725c4949a',1,'dlaf::matrix::createMatrixFromTile(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm, T *ptr)'],['../matrix_8h.html#a6bbd903412fe86fb4e7ad4b9b52a2102',1,'dlaf::matrix::createMatrixFromTile(const GlobalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col, const comm::CommunicatorGrid &comm, const comm::Index2D &source_rank_index, T *ptr)'],['../matrix_8h.html#a71205e9a339140e0f9844c929d963d85',1,'dlaf::matrix::createMatrixFromTile(const GlobalElementSize &size, const TileElementSize &block_size, SizeType ld_tile, SizeType tiles_per_col, const comm::CommunicatorGrid &comm, T *ptr)']]] ]; diff --git a/master/search/functions_3.js b/master/search/functions_3.js index 1d929c2d9f..f723f18c3e 100644 --- a/master/search/functions_3.js +++ b/master/search/functions_3.js @@ -1,16 +1,16 @@ var searchData= [ - ['data_859',['data',['../classdlaf_1_1comm_1_1_message.html#a3684d7d2edac45401785fa9e8561eb73',1,'dlaf::comm::Message::data()'],['../structdlaf_1_1common_1_1_data_descriptor.html#a6f93fef2051c96e6b0100bcb50c280e2',1,'dlaf::common::DataDescriptor::data()']]], - ['data_5fblocksize_860',['data_blocksize',['../data_8h.html#aee766fd59c4a4d3c428082b37b54ea81',1,'dlaf::common']]], - ['data_5fcount_861',['data_count',['../data_8h.html#a4b0d72a2fdf94917b5641c25c86bcf18',1,'dlaf::common']]], - ['data_5fiscontiguous_862',['data_iscontiguous',['../data_8h.html#aee9f8825e02ce8fb557f93bd4d4facef',1,'dlaf::common']]], - ['data_5fnblocks_863',['data_nblocks',['../data_8h.html#a2f5f60205cc53a1a1bce17012cc36368',1,'dlaf::common']]], - ['data_5fpointer_864',['data_pointer',['../data_8h.html#ab4238a013332af8dc361a581d4250205',1,'dlaf::common']]], - ['data_5fstride_865',['data_stride',['../data_8h.html#a18f232284c687293001d172fd19ed3fe',1,'dlaf::common']]], - ['datadescriptor_866',['DataDescriptor',['../structdlaf_1_1common_1_1_data_descriptor.html#a87ca3689fd590f74e636317987af17bd',1,'dlaf::common::DataDescriptor::DataDescriptor()'],['../structdlaf_1_1common_1_1_data_descriptor.html#a8c969a23a32e1cf2ae19a7811ae394e4',1,'dlaf::common::DataDescriptor::DataDescriptor(T *ptr, SizeType n) noexcept'],['../structdlaf_1_1common_1_1_data_descriptor.html#acd73354fd77df45c9b70ab034fc1cd61',1,'dlaf::common::DataDescriptor::DataDescriptor(T *ptr, SizeType num_blocks, SizeType blocksize, SizeType stride) noexcept'],['../structdlaf_1_1common_1_1_data_descriptor.html#acb58ba4559ea0a6022a45bb781dccc6c',1,'dlaf::common::DataDescriptor::DataDescriptor(const Data &data)'],['../structdlaf_1_1common_1_1_data_descriptor_3_01_t_0f_n_0e_4.html#acef674017777de4a735258a476e95e47',1,'dlaf::common::DataDescriptor< T[N]>::DataDescriptor()']]], - ['distancetoadjacenttile_867',['distanceToAdjacentTile',['../classdlaf_1_1matrix_1_1_distribution.html#aa0b67d70b8b3e6f74c53d1062546b8ed',1,'dlaf::matrix::Distribution']]], - ['distribution_868',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html#a47b322556c713fd38223538f50f58e14',1,'dlaf::matrix::Distribution']]], - ['distribution_869',['distribution',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a28967ff509064a6e04cfc8b2d35af402',1,'dlaf::matrix::internal::MatrixBase']]], - ['distribution_870',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html#a012f931d7acdf168a42445bcbbce5dd3',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const TileElementSize &tile_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalTileIndex &tile_offset, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a51c38487398212457417f784dd3d2594',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const TileElementSize &tile_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a97b363d57f32add582ce1774a6b9a147',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a146465cb553b8a9d9f32bf63ff7e7375',1,'dlaf::matrix::Distribution::Distribution(const LocalElementSize &size, const TileElementSize &block_size, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#aacf1f8d3d02d9c70bf0bffe0725b5d92',1,'dlaf::matrix::Distribution::Distribution() noexcept']]], - ['dlaf_5fmake_5fsender_5falgorithm_5foverloads_871',['DLAF_MAKE_SENDER_ALGORITHM_OVERLOADS',['../copy__tile_8h.html#a57758eefb97e5d9fef7405d84c90d867',1,'dlaf::matrix']]] + ['data_861',['data',['../classdlaf_1_1comm_1_1_message.html#a3684d7d2edac45401785fa9e8561eb73',1,'dlaf::comm::Message::data()'],['../structdlaf_1_1common_1_1_data_descriptor.html#a6f93fef2051c96e6b0100bcb50c280e2',1,'dlaf::common::DataDescriptor::data()']]], + ['data_5fblocksize_862',['data_blocksize',['../data_8h.html#aee766fd59c4a4d3c428082b37b54ea81',1,'dlaf::common']]], + ['data_5fcount_863',['data_count',['../data_8h.html#a4b0d72a2fdf94917b5641c25c86bcf18',1,'dlaf::common']]], + ['data_5fiscontiguous_864',['data_iscontiguous',['../data_8h.html#aee9f8825e02ce8fb557f93bd4d4facef',1,'dlaf::common']]], + ['data_5fnblocks_865',['data_nblocks',['../data_8h.html#a2f5f60205cc53a1a1bce17012cc36368',1,'dlaf::common']]], + ['data_5fpointer_866',['data_pointer',['../data_8h.html#ab4238a013332af8dc361a581d4250205',1,'dlaf::common']]], + ['data_5fstride_867',['data_stride',['../data_8h.html#a18f232284c687293001d172fd19ed3fe',1,'dlaf::common']]], + ['datadescriptor_868',['DataDescriptor',['../structdlaf_1_1common_1_1_data_descriptor.html#a87ca3689fd590f74e636317987af17bd',1,'dlaf::common::DataDescriptor::DataDescriptor()'],['../structdlaf_1_1common_1_1_data_descriptor.html#a8c969a23a32e1cf2ae19a7811ae394e4',1,'dlaf::common::DataDescriptor::DataDescriptor(T *ptr, SizeType n) noexcept'],['../structdlaf_1_1common_1_1_data_descriptor.html#acd73354fd77df45c9b70ab034fc1cd61',1,'dlaf::common::DataDescriptor::DataDescriptor(T *ptr, SizeType num_blocks, SizeType blocksize, SizeType stride) noexcept'],['../structdlaf_1_1common_1_1_data_descriptor.html#acb58ba4559ea0a6022a45bb781dccc6c',1,'dlaf::common::DataDescriptor::DataDescriptor(const Data &data)'],['../structdlaf_1_1common_1_1_data_descriptor_3_01_t_0f_n_0e_4.html#acef674017777de4a735258a476e95e47',1,'dlaf::common::DataDescriptor< T[N]>::DataDescriptor()']]], + ['distancetoadjacenttile_869',['distanceToAdjacentTile',['../classdlaf_1_1matrix_1_1_distribution.html#aa0b67d70b8b3e6f74c53d1062546b8ed',1,'dlaf::matrix::Distribution']]], + ['distribution_870',['Distribution',['../classdlaf_1_1matrix_1_1_distribution.html#a47b322556c713fd38223538f50f58e14',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalTileIndex &tile_offset, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a012f931d7acdf168a42445bcbbce5dd3',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const TileElementSize &tile_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalTileIndex &tile_offset, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a51c38487398212457417f784dd3d2594',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const TileElementSize &tile_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a97b363d57f32add582ce1774a6b9a147',1,'dlaf::matrix::Distribution::Distribution(const GlobalElementSize &size, const TileElementSize &block_size, const comm::Size2D &grid_size, const comm::Index2D &rank_index, const comm::Index2D &source_rank_index, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#a146465cb553b8a9d9f32bf63ff7e7375',1,'dlaf::matrix::Distribution::Distribution(const LocalElementSize &size, const TileElementSize &block_size, const GlobalElementIndex &element_offset={0, 0})'],['../classdlaf_1_1matrix_1_1_distribution.html#aacf1f8d3d02d9c70bf0bffe0725b5d92',1,'dlaf::matrix::Distribution::Distribution() noexcept']]], + ['distribution_871',['distribution',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a28967ff509064a6e04cfc8b2d35af402',1,'dlaf::matrix::internal::MatrixBase']]], + ['dlaf_5fmake_5fsender_5falgorithm_5foverloads_872',['DLAF_MAKE_SENDER_ALGORITHM_OVERLOADS',['../copy__tile_8h.html#a57758eefb97e5d9fef7405d84c90d867',1,'dlaf::matrix']]], + ['done_873',['done',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#aaac68e6ec599e12aac6e7e4bd32570a8',1,'dlaf::matrix::Matrix< const T, D >::done(const LocalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#ade093c346a4df7e85c94b71fd86eaec2',1,'dlaf::matrix::Matrix< const T, D >::done(const GlobalTileIndex &index) noexcept']]] ]; diff --git a/master/search/functions_4.js b/master/search/functions_4.js index 268caeee45..685b47bc8f 100644 --- a/master/search/functions_4.js +++ b/master/search/functions_4.js @@ -1,9 +1,9 @@ var searchData= [ - ['eigensolver_872',['eigensolver',['../eigensolver_2eigensolver_8h.html#a7aeed252a1783c24f778eb250d0efd72',1,'dlaf::eigensolver::eigensolver(blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../eigensolver_2eigensolver_8h.html#a94fb08e9087b48ae5fe19b59c9e28db8',1,'dlaf::eigensolver::eigensolver(blas::Uplo uplo, Matrix< T, D > &mat)'],['../eigensolver_2eigensolver_8h.html#a2b99692c7224f63a547cc6686d732490',1,'dlaf::eigensolver::eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../eigensolver_2eigensolver_8h.html#ad43e9ff356aafef70396fdc6b6d3a1f2',1,'dlaf::eigensolver::eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat)']]], - ['elementfromtileandtileelement_873',['elementFromTileAndTileElement',['../util__distribution_8h.html#a9eab8f0da313765c25ddd2eec12665e0',1,'dlaf::util::matrix']]], - ['equal_5fblocksize_874',['equal_blocksize',['../util__matrix_8h.html#a0bb54ace6dfcb245ec387c1af75e63fe',1,'dlaf::matrix']]], - ['equal_5fdistributions_875',['equal_distributions',['../util__matrix_8h.html#a8af8a409c1af4f1c7cdbed9b4f6b24a2',1,'dlaf::matrix']]], - ['equal_5fprocess_5fgrid_876',['equal_process_grid',['../util__matrix_8h.html#a7a41b22af7fe104052668cc02d5a624c',1,'dlaf::matrix']]], - ['equal_5fsize_877',['equal_size',['../util__matrix_8h.html#ada06af4450b10ed70a5830c0d8db0302',1,'dlaf::matrix']]] + ['eigensolver_874',['eigensolver',['../eigensolver_2eigensolver_8h.html#a7aeed252a1783c24f778eb250d0efd72',1,'dlaf::eigensolver::eigensolver(blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../eigensolver_2eigensolver_8h.html#a94fb08e9087b48ae5fe19b59c9e28db8',1,'dlaf::eigensolver::eigensolver(blas::Uplo uplo, Matrix< T, D > &mat)'],['../eigensolver_2eigensolver_8h.html#a2b99692c7224f63a547cc6686d732490',1,'dlaf::eigensolver::eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../eigensolver_2eigensolver_8h.html#ad43e9ff356aafef70396fdc6b6d3a1f2',1,'dlaf::eigensolver::eigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat)']]], + ['elementfromtileandtileelement_875',['elementFromTileAndTileElement',['../util__distribution_8h.html#a9eab8f0da313765c25ddd2eec12665e0',1,'dlaf::util::matrix']]], + ['equal_5fblocksize_876',['equal_blocksize',['../util__matrix_8h.html#a0bb54ace6dfcb245ec387c1af75e63fe',1,'dlaf::matrix']]], + ['equal_5fdistributions_877',['equal_distributions',['../util__matrix_8h.html#a8af8a409c1af4f1c7cdbed9b4f6b24a2',1,'dlaf::matrix']]], + ['equal_5fprocess_5fgrid_878',['equal_process_grid',['../util__matrix_8h.html#a7a41b22af7fe104052668cc02d5a624c',1,'dlaf::matrix']]], + ['equal_5fsize_879',['equal_size',['../util__matrix_8h.html#ada06af4450b10ed70a5830c0d8db0302',1,'dlaf::matrix']]] ]; diff --git a/master/search/functions_5.js b/master/search/functions_5.js index f05df936cb..7fa7712be1 100644 --- a/master/search/functions_5.js +++ b/master/search/functions_5.js @@ -1,6 +1,6 @@ var searchData= [ - ['finalize_878',['finalize',['../init_8h.html#abf5448f6d8cff5d741a7680123a7082c',1,'dlaf']]], - ['fullcommunicator_879',['fullCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#a1cb5842312d3e88e298819ea192d286d',1,'dlaf::comm::CommunicatorGrid']]], - ['fullindex_880',['fullIndex',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a709e5ba882750052d23a5a8eb1eff21a',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] + ['finalize_880',['finalize',['../init_8h.html#abf5448f6d8cff5d741a7680123a7082c',1,'dlaf']]], + ['fullcommunicator_881',['fullCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#a1cb5842312d3e88e298819ea192d286d',1,'dlaf::comm::CommunicatorGrid']]], + ['fullindex_882',['fullIndex',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a709e5ba882750052d23a5a8eb1eff21a',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] ]; diff --git a/master/search/functions_6.js b/master/search/functions_6.js index f3e91804e5..40dacbd343 100644 --- a/master/search/functions_6.js +++ b/master/search/functions_6.js @@ -1,21 +1,21 @@ var searchData= [ - ['gemm_881',['gemm',['../blas_2tile_8h.html#a8e495a0a86ddf2f997ca8a7c33d106a2',1,'dlaf::tile::gemm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#afed112b0e26de470da2469d43866918d',1,'dlaf::tile::gemm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a1de961c6e1c01ee6a5f0039ecf51b162',1,'dlaf::tile::gemm(const blas::Op op_a, const blas::Op op_b, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const T beta, const Tile< T, D > &c)']]], - ['geneigensolver_882',['genEigensolver',['../gen__eigensolver_8h.html#a7578fbe49ba90d7ef916ac81860c3858',1,'dlaf::eigensolver::genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b)'],['../gen__eigensolver_8h.html#a9d80f2453da0de6a6ad73c55165f7f52',1,'dlaf::eigensolver::genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../gen__eigensolver_8h.html#ad60b7fb5135b8ea685d781e4e4165c47',1,'dlaf::eigensolver::genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b)'],['../gen__eigensolver_8h.html#ae119b500260a60dcabdd237e461a12dd',1,'dlaf::eigensolver::genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)']]], - ['generalsubmatrix_883',['generalSubMatrix',['../multiplication_2general_8h.html#a0aa788470aca6ca60202513251578a36',1,'dlaf::multiplication::generalSubMatrix([[maybe_unused]] comm::CommunicatorGrid grid, common::Pipeline< comm::Communicator > &row_task_chain, common::Pipeline< comm::Communicator > &col_task_chain, const SizeType a, const SizeType b, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)'],['../multiplication_2general_8h.html#a45f04fd24259c413a07a210a17b91b93',1,'dlaf::multiplication::generalSubMatrix(const SizeType a, const SizeType b, const blas::Op opA, const blas::Op opB, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)']]], - ['generic_5finteger_5fop_884',['generic_integer_op',['../util__math_8h.html#a4a1c1045d16c92d7fef9c553ae8e6c90',1,'dlaf::util::internal']]], - ['gentostd_885',['genToStd',['../gen__to__std_8h.html#a8f0b0f0094c0185dc428f69f8506ba8f',1,'dlaf::eigensolver::genToStd(blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)'],['../gen__to__std_8h.html#a1c178c1c870afbc1c4928a4dbe078aa4',1,'dlaf::eigensolver::genToStd(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)']]], - ['get_886',['get',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#a8b053672d5ecf63c2af3a3c23677278b',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::get()'],['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a23701a66458fad55204f90b84674a8e2',1,'dlaf::common::internal::basic_coords::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#ad5eaeb945174c2c237a672dd4191582a',1,'dlaf::matrix::MatrixMirror::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#aa0d470acb7c2f97a7192c06fceef7b32',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#ad2cba1bf3133441bc31d960d2681cf11',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::get()']]], - ['getheight_887',['getHeight',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ad59bcb70b44cf2a4a61119d8ce4eb5db',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['getoptionsdescription_888',['getOptionsDescription',['../init_8h.html#a6f4711b407727e4b9317d252b451bd89',1,'dlaf']]], - ['getsource_889',['getSource',['../classdlaf_1_1matrix_1_1_matrix_mirror.html#a2fb277eb0664e0e6ed0e8a011416c7c7',1,'dlaf::matrix::MatrixMirror::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#a424e6b85bb0167015efbb7522a31d7fa',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a348f9f6d9af5df67ef6179e4f600850e',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#af6ab705620e03736d36433f0846470fc',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::getSource()']]], - ['getwidth_890',['getWidth',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa34a2e93ba64d958df32428f3928ccba',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['globalelementfromglobaltileandtileelement_891',['globalElementFromGlobalTileAndTileElement',['../classdlaf_1_1matrix_1_1_distribution.html#aed7a91383a0146b02008c6d6bfed2850',1,'dlaf::matrix::Distribution']]], - ['globalelementfromlocaltileandtileelement_892',['globalElementFromLocalTileAndTileElement',['../classdlaf_1_1matrix_1_1_distribution.html#a36c1a5962e433c773a924d8b5bdda594',1,'dlaf::matrix::Distribution']]], - ['globalelementindex_893',['globalElementIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a20234ff229ec8872625293b28c7d47f5',1,'dlaf::matrix::Distribution']]], - ['globaltileelementdistance_894',['globalTileElementDistance',['../classdlaf_1_1matrix_1_1_distribution.html#a9f57d9074212c024d39f0e971bc9e5db',1,'dlaf::matrix::Distribution']]], - ['globaltilefromglobalelement_895',['globalTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#ae63d662d123b38f1150b553b2cdb259d',1,'dlaf::matrix::Distribution']]], - ['globaltilefromlocaltile_896',['globalTileFromLocalTile',['../util__distribution_8h.html#a3fa80781a6b118e00015b9751e2f6ead',1,'dlaf::util::matrix::globalTileFromLocalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#add4b487f6603fb6099dce376ee2e8c3c',1,'dlaf::matrix::Distribution::globalTileFromLocalTile(SizeType local_tile) const noexcept']]], - ['globaltileindex_897',['globalTileIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a7a0027799c7c187d8f7135a1242ca3f5',1,'dlaf::matrix::Distribution::globalTileIndex(const LocalTileIndex &local_tile) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ae49ef916d77bf8de7e8e4710a1034299',1,'dlaf::matrix::Distribution::globalTileIndex(const GlobalElementIndex &global_element) const noexcept']]], - ['globaltilelinearindex_898',['globalTileLinearIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a0b174a696d5fe710ecb2bc081081c137',1,'dlaf::matrix::Distribution']]] + ['gemm_883',['gemm',['../blas_2tile_8h.html#a8e495a0a86ddf2f997ca8a7c33d106a2',1,'dlaf::tile::gemm(const dlaf::internal::Policy< B > &p)'],['../blas_2tile_8h.html#afed112b0e26de470da2469d43866918d',1,'dlaf::tile::gemm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a1de961c6e1c01ee6a5f0039ecf51b162',1,'dlaf::tile::gemm(const blas::Op op_a, const blas::Op op_b, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const T beta, const Tile< T, D > &c)']]], + ['geneigensolver_884',['genEigensolver',['../gen__eigensolver_8h.html#a7578fbe49ba90d7ef916ac81860c3858',1,'dlaf::eigensolver::genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b)'],['../gen__eigensolver_8h.html#a9d80f2453da0de6a6ad73c55165f7f52',1,'dlaf::eigensolver::genEigensolver(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)'],['../gen__eigensolver_8h.html#ad60b7fb5135b8ea685d781e4e4165c47',1,'dlaf::eigensolver::genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b)'],['../gen__eigensolver_8h.html#ae119b500260a60dcabdd237e461a12dd',1,'dlaf::eigensolver::genEigensolver(blas::Uplo uplo, Matrix< T, D > &mat_a, Matrix< T, D > &mat_b, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)']]], + ['generalsubmatrix_885',['generalSubMatrix',['../multiplication_2general_8h.html#a0aa788470aca6ca60202513251578a36',1,'dlaf::multiplication::generalSubMatrix([[maybe_unused]] comm::CommunicatorGrid grid, common::Pipeline< comm::Communicator > &row_task_chain, common::Pipeline< comm::Communicator > &col_task_chain, const SizeType a, const SizeType b, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)'],['../multiplication_2general_8h.html#a45f04fd24259c413a07a210a17b91b93',1,'dlaf::multiplication::generalSubMatrix(const SizeType a, const SizeType b, const blas::Op opA, const blas::Op opB, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)']]], + ['generic_5finteger_5fop_886',['generic_integer_op',['../util__math_8h.html#a4a1c1045d16c92d7fef9c553ae8e6c90',1,'dlaf::util::internal']]], + ['gentostd_887',['genToStd',['../gen__to__std_8h.html#a8f0b0f0094c0185dc428f69f8506ba8f',1,'dlaf::eigensolver::genToStd(blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)'],['../gen__to__std_8h.html#a1c178c1c870afbc1c4928a4dbe078aa4',1,'dlaf::eigensolver::genToStd(comm::CommunicatorGrid grid, blas::Uplo uplo, Matrix< T, device > &mat_a, Matrix< T, device > &mat_b)']]], + ['get_888',['get',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#a8b053672d5ecf63c2af3a3c23677278b',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::get()'],['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a23701a66458fad55204f90b84674a8e2',1,'dlaf::common::internal::basic_coords::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#ad5eaeb945174c2c237a672dd4191582a',1,'dlaf::matrix::MatrixMirror::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#aa0d470acb7c2f97a7192c06fceef7b32',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::get()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#ad2cba1bf3133441bc31d960d2681cf11',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::get()']]], + ['getheight_889',['getHeight',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ad59bcb70b44cf2a4a61119d8ce4eb5db',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['getoptionsdescription_890',['getOptionsDescription',['../init_8h.html#a6f4711b407727e4b9317d252b451bd89',1,'dlaf']]], + ['getsource_891',['getSource',['../classdlaf_1_1matrix_1_1_matrix_mirror.html#a2fb277eb0664e0e6ed0e8a011416c7c7',1,'dlaf::matrix::MatrixMirror::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#a424e6b85bb0167015efbb7522a31d7fa',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a348f9f6d9af5df67ef6179e4f600850e',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::getSource()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#af6ab705620e03736d36433f0846470fc',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::getSource()']]], + ['getwidth_892',['getWidth',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa34a2e93ba64d958df32428f3928ccba',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['globalelementfromglobaltileandtileelement_893',['globalElementFromGlobalTileAndTileElement',['../classdlaf_1_1matrix_1_1_distribution.html#aed7a91383a0146b02008c6d6bfed2850',1,'dlaf::matrix::Distribution']]], + ['globalelementfromlocaltileandtileelement_894',['globalElementFromLocalTileAndTileElement',['../classdlaf_1_1matrix_1_1_distribution.html#a36c1a5962e433c773a924d8b5bdda594',1,'dlaf::matrix::Distribution']]], + ['globalelementindex_895',['globalElementIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a20234ff229ec8872625293b28c7d47f5',1,'dlaf::matrix::Distribution']]], + ['globaltileelementdistance_896',['globalTileElementDistance',['../classdlaf_1_1matrix_1_1_distribution.html#a9f57d9074212c024d39f0e971bc9e5db',1,'dlaf::matrix::Distribution']]], + ['globaltilefromglobalelement_897',['globalTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#ae63d662d123b38f1150b553b2cdb259d',1,'dlaf::matrix::Distribution']]], + ['globaltilefromlocaltile_898',['globalTileFromLocalTile',['../util__distribution_8h.html#a3fa80781a6b118e00015b9751e2f6ead',1,'dlaf::util::matrix::globalTileFromLocalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#add4b487f6603fb6099dce376ee2e8c3c',1,'dlaf::matrix::Distribution::globalTileFromLocalTile(SizeType local_tile) const noexcept']]], + ['globaltileindex_899',['globalTileIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a7a0027799c7c187d8f7135a1242ca3f5',1,'dlaf::matrix::Distribution::globalTileIndex(const LocalTileIndex &local_tile) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ae49ef916d77bf8de7e8e4710a1034299',1,'dlaf::matrix::Distribution::globalTileIndex(const GlobalElementIndex &global_element) const noexcept']]], + ['globaltilelinearindex_900',['globalTileLinearIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a0b174a696d5fe710ecb2bc081081c137',1,'dlaf::matrix::Distribution']]] ]; diff --git a/master/search/functions_7.js b/master/search/functions_7.js index 7c120c5acf..c18293fbee 100644 --- a/master/search/functions_7.js +++ b/master/search/functions_7.js @@ -1,8 +1,8 @@ var searchData= [ - ['hegst_899',['hegst',['../lapack_2tile_8h.html#aaf274e91e36453e7994dd1f2a2813681',1,'dlaf::tile']]], - ['hemm_900',['hemm',['../blas_2tile_8h.html#aeec422be0ef5abea8ab6f1a8ae40c8ef',1,'dlaf::tile::hemm(const blas::Side side, const blas::Uplo uplo, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const T beta, const Tile< T, D > &c)'],['../blas_2tile_8h.html#ab7031513f6c7338a1f3cea71784d0e9a',1,'dlaf::tile::hemm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a8887bb899e49f4ea3acce81474f5965d',1,'dlaf::tile::hemm(const dlaf::internal::Policy< B > &p)']]], - ['her2k_901',['her2k',['../blas_2tile_8h.html#a36642d0aeb16e71f10fd289985f4a03f',1,'dlaf::tile::her2k(const blas::Uplo uplo, const blas::Op op, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const BaseType< T > beta, const Tile< T, D > &c)'],['../blas_2tile_8h.html#aafcb69907e7154cb12e0e3121534310a',1,'dlaf::tile::her2k(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a45e96a215d6f3a7f64ccf05d108087e7',1,'dlaf::tile::her2k(const dlaf::internal::Policy< B > &p)']]], - ['herk_902',['herk',['../blas_2tile_8h.html#a72ab20a0ba5ba199c1ec18f8a31ee2eb',1,'dlaf::tile::herk(const blas::Uplo uplo, const blas::Op op, const BaseType< T > alpha, const Tile< const T, D > &a, const BaseType< T > beta, const Tile< T, D > &c)'],['../blas_2tile_8h.html#a8a4f39f0b26f02bc094e531add909e6e',1,'dlaf::tile::herk(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#aa8fd52359e36023fb6418834d4b27fcd',1,'dlaf::tile::herk(const dlaf::internal::Policy< B > &p)']]], - ['hermitian_903',['hermitian',['../hermitian_8h.html#a117a0b955e26513fecdc562da0f96922',1,'dlaf::multiplication::hermitian(blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)'],['../hermitian_8h.html#a3a63ff51bec2e9383f3dfd8b2914673e',1,'dlaf::multiplication::hermitian(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)']]] + ['hegst_901',['hegst',['../lapack_2tile_8h.html#aaf274e91e36453e7994dd1f2a2813681',1,'dlaf::tile']]], + ['hemm_902',['hemm',['../blas_2tile_8h.html#aeec422be0ef5abea8ab6f1a8ae40c8ef',1,'dlaf::tile::hemm(const blas::Side side, const blas::Uplo uplo, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const T beta, const Tile< T, D > &c)'],['../blas_2tile_8h.html#ab7031513f6c7338a1f3cea71784d0e9a',1,'dlaf::tile::hemm(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a8887bb899e49f4ea3acce81474f5965d',1,'dlaf::tile::hemm(const dlaf::internal::Policy< B > &p)']]], + ['her2k_903',['her2k',['../blas_2tile_8h.html#a36642d0aeb16e71f10fd289985f4a03f',1,'dlaf::tile::her2k(const blas::Uplo uplo, const blas::Op op, const T alpha, const Tile< const T, D > &a, const Tile< const T, D > &b, const BaseType< T > beta, const Tile< T, D > &c)'],['../blas_2tile_8h.html#aafcb69907e7154cb12e0e3121534310a',1,'dlaf::tile::her2k(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#a45e96a215d6f3a7f64ccf05d108087e7',1,'dlaf::tile::her2k(const dlaf::internal::Policy< B > &p)']]], + ['herk_904',['herk',['../blas_2tile_8h.html#a72ab20a0ba5ba199c1ec18f8a31ee2eb',1,'dlaf::tile::herk(const blas::Uplo uplo, const blas::Op op, const BaseType< T > alpha, const Tile< const T, D > &a, const BaseType< T > beta, const Tile< T, D > &c)'],['../blas_2tile_8h.html#a8a4f39f0b26f02bc094e531add909e6e',1,'dlaf::tile::herk(const dlaf::internal::Policy< B > &p, Sender &&s)'],['../blas_2tile_8h.html#aa8fd52359e36023fb6418834d4b27fcd',1,'dlaf::tile::herk(const dlaf::internal::Policy< B > &p)']]], + ['hermitian_905',['hermitian',['../hermitian_8h.html#a117a0b955e26513fecdc562da0f96922',1,'dlaf::multiplication::hermitian(blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)'],['../hermitian_8h.html#a3a63ff51bec2e9383f3dfd8b2914673e',1,'dlaf::multiplication::hermitian(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, const T alpha, Matrix< const T, D > &mat_a, Matrix< const T, D > &mat_b, const T beta, Matrix< T, D > &mat_c)']]] ]; diff --git a/master/search/functions_8.js b/master/search/functions_8.js index 8861c1cd9c..ed957d3c7c 100644 --- a/master/search/functions_8.js +++ b/master/search/functions_8.js @@ -1,14 +1,14 @@ var searchData= [ - ['index2d_904',['Index2D',['../classdlaf_1_1common_1_1_index2_d.html#a940432586f1b501b8dda46355999335d',1,'dlaf::common::Index2D::Index2D() noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#ae959b1b0c405d3480f9d4960946b5b2e',1,'dlaf::common::Index2D::Index2D(const std::array< IndexT, 2 > &coords) noexcept']]], - ['initialize_905',['initialize',['../init_8h.html#a100757e13ff576c8a93dd7aed3ca7ec4',1,'dlaf::initialize(const pika::program_options::variables_map &vm, const configuration &user_cfg={})'],['../init_8h.html#a94175c61baedbcc0bf6bd53e0f57380f',1,'dlaf::initialize(int argc, const char *const argv[], const configuration &user_cfg={})']]], - ['initresourcepartitionerhandler_906',['initResourcePartitionerHandler',['../init_8h.html#a3685b1a9ef2da35287ae3b03ea3b62e6',1,'dlaf']]], - ['is_5fcontiguous_907',['is_contiguous',['../structdlaf_1_1common_1_1_data_descriptor.html#ad73cebe7217dd9f3ad68190dc09bce9e',1,'dlaf::common::DataDescriptor']]], - ['isempty_908',['isEmpty',['../classdlaf_1_1common_1_1_size2_d.html#afe5870841ddc785db5d4c81cc22aeb93',1,'dlaf::common::Size2D']]], - ['isexternal_909',['isExternal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa77cf35507fc6a6291e9197996034db8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['isin_910',['isIn',['../classdlaf_1_1common_1_1_index2_d.html#a5f174cc234af20d79923b31d7fa90b19',1,'dlaf::common::Index2D']]], - ['isinoron_911',['isInOrOn',['../classdlaf_1_1common_1_1_index2_d.html#ab230f7ada23f15ffd3b3d0a3921f1685',1,'dlaf::common::Index2D']]], - ['isvalid_912',['isValid',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a380f1ac07da35bc014bcbfff908eb0a9',1,'dlaf::common::internal::basic_coords']]], - ['iterate_5frange2d_913',['iterate_range2d',['../range2d_8h.html#adcdcf40345b7f03cf52e871a2420057b',1,'dlaf::common']]], - ['iteratorlocal_914',['iteratorLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a79626e0d68d508aca69d1114cedc38a7',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::iteratorLocal()'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7c25686d3fb7e47b627af360592168ad',1,'dlaf::matrix::internal::View::iteratorLocal()']]] + ['index2d_906',['Index2D',['../classdlaf_1_1common_1_1_index2_d.html#a940432586f1b501b8dda46355999335d',1,'dlaf::common::Index2D::Index2D() noexcept'],['../classdlaf_1_1common_1_1_index2_d.html#ae959b1b0c405d3480f9d4960946b5b2e',1,'dlaf::common::Index2D::Index2D(const std::array< IndexT, 2 > &coords) noexcept']]], + ['initialize_907',['initialize',['../init_8h.html#a100757e13ff576c8a93dd7aed3ca7ec4',1,'dlaf::initialize(const pika::program_options::variables_map &vm, const configuration &user_cfg={})'],['../init_8h.html#a94175c61baedbcc0bf6bd53e0f57380f',1,'dlaf::initialize(int argc, const char *const argv[], const configuration &user_cfg={})']]], + ['initresourcepartitionerhandler_908',['initResourcePartitionerHandler',['../init_8h.html#a3685b1a9ef2da35287ae3b03ea3b62e6',1,'dlaf']]], + ['is_5fcontiguous_909',['is_contiguous',['../structdlaf_1_1common_1_1_data_descriptor.html#ad73cebe7217dd9f3ad68190dc09bce9e',1,'dlaf::common::DataDescriptor']]], + ['isempty_910',['isEmpty',['../classdlaf_1_1common_1_1_size2_d.html#afe5870841ddc785db5d4c81cc22aeb93',1,'dlaf::common::Size2D']]], + ['isexternal_911',['isExternal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa77cf35507fc6a6291e9197996034db8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['isin_912',['isIn',['../classdlaf_1_1common_1_1_index2_d.html#a5f174cc234af20d79923b31d7fa90b19',1,'dlaf::common::Index2D']]], + ['isinoron_913',['isInOrOn',['../classdlaf_1_1common_1_1_index2_d.html#ab230f7ada23f15ffd3b3d0a3921f1685',1,'dlaf::common::Index2D']]], + ['isvalid_914',['isValid',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a380f1ac07da35bc014bcbfff908eb0a9',1,'dlaf::common::internal::basic_coords']]], + ['iterate_5frange2d_915',['iterate_range2d',['../range2d_8h.html#adcdcf40345b7f03cf52e871a2420057b',1,'dlaf::common']]], + ['iteratorlocal_916',['iteratorLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a79626e0d68d508aca69d1114cedc38a7',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::iteratorLocal()'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7c25686d3fb7e47b627af360592168ad',1,'dlaf::matrix::internal::View::iteratorLocal()']]] ]; diff --git a/master/search/functions_9.js b/master/search/functions_9.js index cc26976aff..833905c7cf 100644 --- a/master/search/functions_9.js +++ b/master/search/functions_9.js @@ -1,18 +1,18 @@ var searchData= [ - ['lacpy_915',['lacpy',['../util__lapack_8h.html#a2a62e0ec6768b42ae19efff5ea5e4331',1,'lapack::lacpy()'],['../lapack_2tile_8h.html#a1bbf24144aebc55d9d50bb3b53ae38f7',1,'dlaf::tile::lacpy(TileElementSize region, TileElementIndex in_idx, const Tile< const T, Device::CPU > &in, TileElementIndex out_idx, const Tile< T, Device::CPU > &out)'],['../lapack_2tile_8h.html#adba78fac5535c410d8bd2e6c53ff8c7d',1,'dlaf::tile::lacpy(const Tile< const T, Device::CPU > &a, const Tile< T, Device::CPU > &b)']]], - ['lange_916',['lange',['../lapack_2tile_8h.html#a839d86dad6ce772207964d76b3ad07fa',1,'dlaf::tile']]], - ['lantr_917',['lantr',['../lapack_2tile_8h.html#a22be986552edcbdc4a3785a72620d0ca',1,'dlaf::tile']]], - ['laset_918',['laset',['../lapack_2tile_8h.html#a19439efac55cc846b684ed2c6f78eb33',1,'dlaf::tile::laset()'],['../util__lapack_8h.html#a653b4d466d700c7d6d3befb51c8ca9d8',1,'lapack::laset()']]], - ['layoutinfo_919',['LayoutInfo',['../classdlaf_1_1matrix_1_1_layout_info.html#a8c41196b99f84410426dfdd62609b982',1,'dlaf::matrix::LayoutInfo']]], - ['ld_920',['ld',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a44282c10a5112a78c95c557538f13a74',1,'dlaf::matrix::Tile< const T, D >']]], - ['linearindex_921',['linearIndex',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a1d863b84e40d0567f731413ff788a1a6',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['local_5fmatrix_922',['local_matrix',['../util__matrix_8h.html#a801d4c1863ecd48a5758741b63781921',1,'dlaf::matrix']]], - ['localelementdistancefromglobaltile_923',['localElementDistanceFromGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#a532cc910ad683b4bf2154e5b01540e61',1,'dlaf::matrix::Distribution::localElementDistanceFromGlobalTile(GlobalTileIndex begin, GlobalTileIndex end) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ac0eef1aae43a7fc895cd280cac21eaa7',1,'dlaf::matrix::Distribution::localElementDistanceFromGlobalTile(SizeType i_begin, SizeType i_end) const noexcept']]], - ['localelementdistancefromlocaltile_924',['localElementDistanceFromLocalTile',['../classdlaf_1_1matrix_1_1_distribution.html#ab590472ea7c2132b30b725954f8ea3c0',1,'dlaf::matrix::Distribution::localElementDistanceFromLocalTile(LocalTileIndex begin, LocalTileIndex end) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#a757d638249f12dd13ffbaa95a9aba21a',1,'dlaf::matrix::Distribution::localElementDistanceFromLocalTile(SizeType i_loc_begin, SizeType i_loc_end) const noexcept']]], - ['localnrtiles_925',['localNrTiles',['../classdlaf_1_1matrix_1_1_distribution.html#a6e8c79e3f901845182c4e7e32de5eb34',1,'dlaf::matrix::Distribution']]], - ['localtilefromglobalelement_926',['localTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#ae812ba2b5ba6b753d733abb48acc5d72',1,'dlaf::matrix::Distribution']]], - ['localtilefromglobaltile_927',['localTileFromGlobalTile',['../util__distribution_8h.html#a235a8d1c90d534e32f93f2ce0e364700',1,'dlaf::util::matrix::localTileFromGlobalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#abdc93bce29920c2dff64235391cf3811',1,'dlaf::matrix::Distribution::localTileFromGlobalTile(SizeType global_tile) const noexcept']]], - ['localtileindex_928',['localTileIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a77f2413076c87fee45f6afb6e6e1baae',1,'dlaf::matrix::Distribution']]], - ['localtilelinearindex_929',['localTileLinearIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a22259431b76866f2f6acfe1a47ab3ace',1,'dlaf::matrix::Distribution']]] + ['lacpy_917',['lacpy',['../util__lapack_8h.html#a2a62e0ec6768b42ae19efff5ea5e4331',1,'lapack::lacpy()'],['../lapack_2tile_8h.html#a1bbf24144aebc55d9d50bb3b53ae38f7',1,'dlaf::tile::lacpy(TileElementSize region, TileElementIndex in_idx, const Tile< const T, Device::CPU > &in, TileElementIndex out_idx, const Tile< T, Device::CPU > &out)'],['../lapack_2tile_8h.html#adba78fac5535c410d8bd2e6c53ff8c7d',1,'dlaf::tile::lacpy(const Tile< const T, Device::CPU > &a, const Tile< T, Device::CPU > &b)']]], + ['lange_918',['lange',['../lapack_2tile_8h.html#a839d86dad6ce772207964d76b3ad07fa',1,'dlaf::tile']]], + ['lantr_919',['lantr',['../lapack_2tile_8h.html#a22be986552edcbdc4a3785a72620d0ca',1,'dlaf::tile']]], + ['laset_920',['laset',['../lapack_2tile_8h.html#a19439efac55cc846b684ed2c6f78eb33',1,'dlaf::tile::laset()'],['../util__lapack_8h.html#a653b4d466d700c7d6d3befb51c8ca9d8',1,'lapack::laset()']]], + ['layoutinfo_921',['LayoutInfo',['../classdlaf_1_1matrix_1_1_layout_info.html#a8c41196b99f84410426dfdd62609b982',1,'dlaf::matrix::LayoutInfo']]], + ['ld_922',['ld',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a44282c10a5112a78c95c557538f13a74',1,'dlaf::matrix::Tile< const T, D >']]], + ['linearindex_923',['linearIndex',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a1d863b84e40d0567f731413ff788a1a6',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['local_5fmatrix_924',['local_matrix',['../util__matrix_8h.html#a801d4c1863ecd48a5758741b63781921',1,'dlaf::matrix']]], + ['localelementdistancefromglobaltile_925',['localElementDistanceFromGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#a532cc910ad683b4bf2154e5b01540e61',1,'dlaf::matrix::Distribution::localElementDistanceFromGlobalTile(GlobalTileIndex begin, GlobalTileIndex end) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ac0eef1aae43a7fc895cd280cac21eaa7',1,'dlaf::matrix::Distribution::localElementDistanceFromGlobalTile(SizeType i_begin, SizeType i_end) const noexcept']]], + ['localelementdistancefromlocaltile_926',['localElementDistanceFromLocalTile',['../classdlaf_1_1matrix_1_1_distribution.html#ab590472ea7c2132b30b725954f8ea3c0',1,'dlaf::matrix::Distribution::localElementDistanceFromLocalTile(LocalTileIndex begin, LocalTileIndex end) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#a757d638249f12dd13ffbaa95a9aba21a',1,'dlaf::matrix::Distribution::localElementDistanceFromLocalTile(SizeType i_loc_begin, SizeType i_loc_end) const noexcept']]], + ['localnrtiles_927',['localNrTiles',['../classdlaf_1_1matrix_1_1_distribution.html#a6e8c79e3f901845182c4e7e32de5eb34',1,'dlaf::matrix::Distribution']]], + ['localtilefromglobalelement_928',['localTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#ae812ba2b5ba6b753d733abb48acc5d72',1,'dlaf::matrix::Distribution']]], + ['localtilefromglobaltile_929',['localTileFromGlobalTile',['../util__distribution_8h.html#a235a8d1c90d534e32f93f2ce0e364700',1,'dlaf::util::matrix::localTileFromGlobalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#abdc93bce29920c2dff64235391cf3811',1,'dlaf::matrix::Distribution::localTileFromGlobalTile(SizeType global_tile) const noexcept']]], + ['localtileindex_930',['localTileIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a77f2413076c87fee45f6afb6e6e1baae',1,'dlaf::matrix::Distribution']]], + ['localtilelinearindex_931',['localTileLinearIndex',['../classdlaf_1_1matrix_1_1_distribution.html#a22259431b76866f2f6acfe1a47ab3ace',1,'dlaf::matrix::Distribution']]] ]; diff --git a/master/search/functions_a.js b/master/search/functions_a.js index 41e3e0d5d1..163a312d0c 100644 --- a/master/search/functions_a.js +++ b/master/search/functions_a.js @@ -1,20 +1,20 @@ var searchData= [ - ['make_5fcommunicator_5fmanaged_930',['make_communicator_managed',['../communicator_8h.html#a8b7bf4fa5dfb408439c9c93ce3ebb216',1,'dlaf::comm']]], - ['make_5fcontiguous_931',['make_contiguous',['../data__descriptor_8h.html#a96a67d2223bad7ea28fcca7041877560',1,'dlaf::common']]], - ['make_5fdata_932',['make_data',['../data_8h.html#ae8a4cf6d87f0a72e5daad8a35478f051',1,'dlaf::common::make_data(Data &&data) noexcept'],['../data_8h.html#aa3aa1962f3a8382cbf457db22c8ca48f',1,'dlaf::common::make_data(T &&data, Ts &&... args) noexcept']]], - ['make_5fmessage_933',['make_message',['../message_8h.html#a256282d96e42c4ff2e6353ae28263422',1,'dlaf::comm']]], - ['matrix_934',['Matrix',['../classdlaf_1_1matrix_1_1_matrix.html#a65b04ca6f55cf213c52cda2f3a532023',1,'dlaf::matrix::Matrix::Matrix(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm)'],['../classdlaf_1_1matrix_1_1_matrix.html#aa0614a38addeb9aaf83d268f1092b779',1,'dlaf::matrix::Matrix::Matrix(const LocalElementSize &size, const TileElementSize &block_size)'],['../classdlaf_1_1matrix_1_1_matrix.html#a5d2827ca2dbad4d6e6482e4c1e963472',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution, const LayoutInfo &layout, ElementType *ptr) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#ad5f429b579c79194b453206f24d7de49',1,'dlaf::matrix::Matrix::Matrix(const LayoutInfo &layout, ElementType *ptr)'],['../classdlaf_1_1matrix_1_1_matrix.html#a1f0c98f29fa14cccef9f5378f6ee20b1',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution, const LayoutInfo &layout) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#a73b38df3414aa5272abd3b6e87d67e37',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution)']]], - ['matrixmirror_935',['MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a66b7ac6e591f693cd7bb8119955055fc',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#ac9dfb2578a5c94b8cf6bf0d5b3186979',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#a355ab06273bb75ba0fb40119829d2aa0',1,'dlaf::matrix::MatrixMirror::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#ab4b4d7a17b5f0ae45d301ef415f41b29',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::MatrixMirror()']]], - ['memorychunk_936',['MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html#a4059fde641a91f8cac05721fb54c749b',1,'dlaf::memory::MemoryChunk::MemoryChunk()'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a53cffddd0ddfd2e1e0db839157a58913',1,'dlaf::memory::MemoryChunk::MemoryChunk(SizeType size)'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a8855d0d0e0c35c83394ce82cd14ea36d',1,'dlaf::memory::MemoryChunk::MemoryChunk(T *ptr, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_chunk.html#ab7a334b89c214fe25ffde4bf0fd4e802',1,'dlaf::memory::MemoryChunk::MemoryChunk(MemoryChunk &&rhs) noexcept']]], - ['memoryview_937',['MemoryView',['../classdlaf_1_1memory_1_1_memory_view.html#a76c5e9d28626abb2991dd6ad1ce460c7',1,'dlaf::memory::MemoryView::MemoryView(const MemoryView &memory_view, SizeType offset, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#a0835f719277ae4f7b97c68d097a07071',1,'dlaf::memory::MemoryView::MemoryView(T *ptr, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#a0650b61938741469c7583d671240619d',1,'dlaf::memory::MemoryView::MemoryView(SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#aa0752bb38aa279bd6260e2d440b8ca7d',1,'dlaf::memory::MemoryView::MemoryView()']]], - ['message_938',['Message',['../classdlaf_1_1comm_1_1_message.html#aa0a15382490f3f628d14f734aeddf285',1,'dlaf::comm::Message']]], - ['minmemsize_939',['minMemSize',['../classdlaf_1_1matrix_1_1_layout_info.html#ad51fa263d160dce4f4db2943720d2506',1,'dlaf::matrix::LayoutInfo']]], - ['mintilememsize_940',['minTileMemSize',['../classdlaf_1_1matrix_1_1_layout_info.html#a90af3a4b4a456d3107dec3eb4c207c29',1,'dlaf::matrix::LayoutInfo::minTileMemSize(const LocalTileIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_layout_info.html#aa13f1acc6c4df1d0e0c11204d3c81ba0',1,'dlaf::matrix::LayoutInfo::minTileMemSize(const TileElementSize &tile_size) const noexcept']]], - ['mpi_5fcall_941',['mpi_call',['../communication_2error_8h.html#aa640fad0edb091d239f68708fbe00103',1,'dlaf::internal']]], - ['mpi_5finit_942',['mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html#a415e55c02e205e6f782ff7000755c842',1,'dlaf::comm::mpi_init']]], - ['mpi_5ftype_943',['mpi_type',['../classdlaf_1_1comm_1_1_message.html#ac9fa892f3513f9b60ad178d1f788f0af',1,'dlaf::comm::Message']]], - ['mul_944',['mul',['../util__math_8h.html#a77a106250e33d119c5415546f6b74544',1,'dlaf::util::size_t::mul()'],['../util__math_8h.html#a090c5b89334c2c21574610d8f0881b67',1,'dlaf::util::ptrdiff_t::mul()']]], - ['multipliable_945',['multipliable',['../util__matrix_8h.html#a14086310f58c9ace5ceb41e040277c86',1,'dlaf::matrix']]], - ['multipliable_5fsizes_946',['multipliable_sizes',['../util__matrix_8h.html#a93b53e9c0e942a5caa96a24242864400',1,'dlaf::matrix']]] + ['make_5fcommunicator_5fmanaged_932',['make_communicator_managed',['../communicator_8h.html#a8b7bf4fa5dfb408439c9c93ce3ebb216',1,'dlaf::comm']]], + ['make_5fcontiguous_933',['make_contiguous',['../data__descriptor_8h.html#a96a67d2223bad7ea28fcca7041877560',1,'dlaf::common']]], + ['make_5fdata_934',['make_data',['../data_8h.html#ae8a4cf6d87f0a72e5daad8a35478f051',1,'dlaf::common::make_data(Data &&data) noexcept'],['../data_8h.html#aa3aa1962f3a8382cbf457db22c8ca48f',1,'dlaf::common::make_data(T &&data, Ts &&... args) noexcept']]], + ['make_5fmessage_935',['make_message',['../message_8h.html#a256282d96e42c4ff2e6353ae28263422',1,'dlaf::comm']]], + ['matrix_936',['Matrix',['../classdlaf_1_1matrix_1_1_matrix.html#a65b04ca6f55cf213c52cda2f3a532023',1,'dlaf::matrix::Matrix::Matrix(const GlobalElementSize &size, const TileElementSize &block_size, const comm::CommunicatorGrid &comm)'],['../classdlaf_1_1matrix_1_1_matrix.html#aa0614a38addeb9aaf83d268f1092b779',1,'dlaf::matrix::Matrix::Matrix(const LocalElementSize &size, const TileElementSize &block_size)'],['../classdlaf_1_1matrix_1_1_matrix.html#a5d2827ca2dbad4d6e6482e4c1e963472',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution, const LayoutInfo &layout, ElementType *ptr) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#ad5f429b579c79194b453206f24d7de49',1,'dlaf::matrix::Matrix::Matrix(const LayoutInfo &layout, ElementType *ptr)'],['../classdlaf_1_1matrix_1_1_matrix.html#a1f0c98f29fa14cccef9f5378f6ee20b1',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution, const LayoutInfo &layout) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#a73b38df3414aa5272abd3b6e87d67e37',1,'dlaf::matrix::Matrix::Matrix(Distribution distribution)']]], + ['matrixmirror_937',['MatrixMirror',['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01_t_00_01_source_target_00_01_source_target_01_4.html#a66b7ac6e591f693cd7bb8119955055fc',1,'dlaf::matrix::MatrixMirror< T, SourceTarget, SourceTarget >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_target_00_01_source_01_4.html#ac9dfb2578a5c94b8cf6bf0d5b3186979',1,'dlaf::matrix::MatrixMirror< const T, Target, Source >::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror.html#a355ab06273bb75ba0fb40119829d2aa0',1,'dlaf::matrix::MatrixMirror::MatrixMirror()'],['../classdlaf_1_1matrix_1_1_matrix_mirror_3_01const_01_t_00_01_source_target_00_01_source_target_01_4.html#ab4b4d7a17b5f0ae45d301ef415f41b29',1,'dlaf::matrix::MatrixMirror< const T, SourceTarget, SourceTarget >::MatrixMirror()']]], + ['memorychunk_938',['MemoryChunk',['../classdlaf_1_1memory_1_1_memory_chunk.html#a4059fde641a91f8cac05721fb54c749b',1,'dlaf::memory::MemoryChunk::MemoryChunk()'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a53cffddd0ddfd2e1e0db839157a58913',1,'dlaf::memory::MemoryChunk::MemoryChunk(SizeType size)'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a8855d0d0e0c35c83394ce82cd14ea36d',1,'dlaf::memory::MemoryChunk::MemoryChunk(T *ptr, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_chunk.html#ab7a334b89c214fe25ffde4bf0fd4e802',1,'dlaf::memory::MemoryChunk::MemoryChunk(MemoryChunk &&rhs) noexcept']]], + ['memoryview_939',['MemoryView',['../classdlaf_1_1memory_1_1_memory_view.html#a76c5e9d28626abb2991dd6ad1ce460c7',1,'dlaf::memory::MemoryView::MemoryView(const MemoryView &memory_view, SizeType offset, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#a0835f719277ae4f7b97c68d097a07071',1,'dlaf::memory::MemoryView::MemoryView(T *ptr, SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#a0650b61938741469c7583d671240619d',1,'dlaf::memory::MemoryView::MemoryView(SizeType size)'],['../classdlaf_1_1memory_1_1_memory_view.html#aa0752bb38aa279bd6260e2d440b8ca7d',1,'dlaf::memory::MemoryView::MemoryView()']]], + ['message_940',['Message',['../classdlaf_1_1comm_1_1_message.html#aa0a15382490f3f628d14f734aeddf285',1,'dlaf::comm::Message']]], + ['minmemsize_941',['minMemSize',['../classdlaf_1_1matrix_1_1_layout_info.html#ad51fa263d160dce4f4db2943720d2506',1,'dlaf::matrix::LayoutInfo']]], + ['mintilememsize_942',['minTileMemSize',['../classdlaf_1_1matrix_1_1_layout_info.html#a90af3a4b4a456d3107dec3eb4c207c29',1,'dlaf::matrix::LayoutInfo::minTileMemSize(const LocalTileIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_layout_info.html#aa13f1acc6c4df1d0e0c11204d3c81ba0',1,'dlaf::matrix::LayoutInfo::minTileMemSize(const TileElementSize &tile_size) const noexcept']]], + ['mpi_5fcall_943',['mpi_call',['../communication_2error_8h.html#aa640fad0edb091d239f68708fbe00103',1,'dlaf::internal']]], + ['mpi_5finit_944',['mpi_init',['../structdlaf_1_1comm_1_1mpi__init.html#a415e55c02e205e6f782ff7000755c842',1,'dlaf::comm::mpi_init']]], + ['mpi_5ftype_945',['mpi_type',['../classdlaf_1_1comm_1_1_message.html#ac9fa892f3513f9b60ad178d1f788f0af',1,'dlaf::comm::Message']]], + ['mul_946',['mul',['../util__math_8h.html#a77a106250e33d119c5415546f6b74544',1,'dlaf::util::size_t::mul()'],['../util__math_8h.html#a090c5b89334c2c21574610d8f0881b67',1,'dlaf::util::ptrdiff_t::mul()']]], + ['multipliable_947',['multipliable',['../util__matrix_8h.html#a14086310f58c9ace5ceb41e040277c86',1,'dlaf::matrix']]], + ['multipliable_5fsizes_948',['multipliable_sizes',['../util__matrix_8h.html#a93b53e9c0e942a5caa96a24242864400',1,'dlaf::matrix']]] ]; diff --git a/master/search/functions_b.js b/master/search/functions_b.js index fe0cea232b..39368903d2 100644 --- a/master/search/functions_b.js +++ b/master/search/functions_b.js @@ -1,8 +1,8 @@ var searchData= [ - ['nblocks_947',['nblocks',['../structdlaf_1_1common_1_1_data_descriptor.html#a583161cb3c61e3af88fe66552a123866',1,'dlaf::common::DataDescriptor']]], - ['nextlocaltilefromglobalelement_948',['nextLocalTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a559caf9af0f658507a6a8fe949e09a93',1,'dlaf::matrix::Distribution']]], - ['nextlocaltilefromglobaltile_949',['nextLocalTileFromGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#afcd76386ddd1319978242aaf54412825',1,'dlaf::matrix::Distribution::nextLocalTileFromGlobalTile()'],['../util__distribution_8h.html#ad6d38066f4c7c00e41fc2d8830efb77f',1,'dlaf::util::matrix::nextLocalTileFromGlobalTile()']]], - ['norm_950',['norm',['../norm_8h.html#a2d0adfb7be9cfe934f96e9dd1c38d839',1,'dlaf::auxiliary']]], - ['nrtiles_951',['nrTiles',['../classdlaf_1_1matrix_1_1_distribution.html#acc85a62e6350bcac93a8e442d4603516',1,'dlaf::matrix::Distribution::nrTiles()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a068e58f7f6c2956f92e69fd7c8d8f68d',1,'dlaf::matrix::internal::MatrixBase::nrTiles()']]] + ['nblocks_949',['nblocks',['../structdlaf_1_1common_1_1_data_descriptor.html#a583161cb3c61e3af88fe66552a123866',1,'dlaf::common::DataDescriptor']]], + ['nextlocaltilefromglobalelement_950',['nextLocalTileFromGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a559caf9af0f658507a6a8fe949e09a93',1,'dlaf::matrix::Distribution']]], + ['nextlocaltilefromglobaltile_951',['nextLocalTileFromGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#afcd76386ddd1319978242aaf54412825',1,'dlaf::matrix::Distribution::nextLocalTileFromGlobalTile()'],['../util__distribution_8h.html#ad6d38066f4c7c00e41fc2d8830efb77f',1,'dlaf::util::matrix::nextLocalTileFromGlobalTile()']]], + ['norm_952',['norm',['../norm_8h.html#a2d0adfb7be9cfe934f96e9dd1c38d839',1,'dlaf::auxiliary']]], + ['nrtiles_953',['nrTiles',['../classdlaf_1_1matrix_1_1_distribution.html#acc85a62e6350bcac93a8e442d4603516',1,'dlaf::matrix::Distribution::nrTiles()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a068e58f7f6c2956f92e69fd7c8d8f68d',1,'dlaf::matrix::internal::MatrixBase::nrTiles()']]] ]; diff --git a/master/search/functions_c.js b/master/search/functions_c.js index a8468d79fb..d201232b5a 100644 --- a/master/search/functions_c.js +++ b/master/search/functions_c.js @@ -1,16 +1,16 @@ var searchData= [ - ['offset_952',['offset',['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7229ff87452798d32b7195fe684fff7a',1,'dlaf::matrix::internal::View']]], - ['offsetelement_953',['offsetElement',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a398b31442f71b68e39a3ecbb2762203e',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::offsetElement()'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7164cccc46c63cc96661288809a37f5f',1,'dlaf::matrix::internal::View::offsetElement()']]], - ['operator_20bool_954',['operator bool',['../structdlaf_1_1common_1_1_buffer.html#ae57231c34ebf6f79da4d1e6e1c75cf0a',1,'dlaf::common::Buffer::operator bool()'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a8f6d65a5adc84620ed33a7de52bac602',1,'dlaf::comm::internal::type_handler::operator bool()']]], - ['operator_20mpi_5fcomm_955',['operator MPI_Comm',['../classdlaf_1_1comm_1_1_communicator.html#a5be809bbc2d038669e31a9de501a2070',1,'dlaf::comm::Communicator']]], - ['operator_20mpi_5fdatatype_956',['operator MPI_Datatype',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a2247361b0d7dd77039860c0cd21a13e3',1,'dlaf::comm::internal::type_handler']]], - ['operator_21_3d_957',['operator!=',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#abf2886043fbe929f1e1f0eff5e20935d',1,'dlaf::common::internal::basic_coords::operator!=()'],['../classdlaf_1_1common_1_1_size2_d.html#a1a6822fefe13303612ce0ddf44f260ea',1,'dlaf::common::Size2D::operator!=()'],['../classdlaf_1_1common_1_1_index2_d.html#ae2e28cf6eab4febd97d6c46d23c6096d',1,'dlaf::common::Index2D::operator!=()']]], - ['operator_26_958',['operator&',['../classdlaf_1_1comm_1_1_communicator.html#ace61947783c442bf684c2f035093acf2',1,'dlaf::comm::Communicator']]], - ['operator_28_29_959',['operator()',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#ac22706782f9864d6357df2f48f15ee74',1,'dlaf::matrix::Tile< const T, D >::operator()()'],['../classdlaf_1_1memory_1_1_memory_view.html#aad118548f59d6c8021eb4fbe816a1561',1,'dlaf::memory::MemoryView::operator()() const'],['../classdlaf_1_1memory_1_1_memory_view.html#a6f9c8123e686e2f2122ec988ca0b9725',1,'dlaf::memory::MemoryView::operator()(SizeType index) const'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a1720c41c3c3cbad2e198e7cedbfa84de',1,'dlaf::memory::MemoryChunk::operator()()'],['../classdlaf_1_1memory_1_1_memory_chunk.html#ac0325c028366b219597dde97814e9a7a',1,'dlaf::memory::MemoryChunk::operator()(SizeType index)'],['../classdlaf_1_1matrix_1_1_tile.html#a5641fc26b7eb759a31ddc24975393ce7',1,'dlaf::matrix::Tile::operator()()'],['../classdlaf_1_1common_1_1_pipeline.html#a388b1fe53e84dde5b0f026e7b334902d',1,'dlaf::common::Pipeline::operator()()']]], - ['operator_2b_960',['operator+',['../index2d_8h.html#a2ed7f0c019788f600bc067956646bf75',1,'dlaf::common']]], - ['operator_3c_3c_961',['operator<<',['../index2d_8h.html#a352cdd70886a2a2ace1abd246951c538',1,'dlaf::common']]], - ['operator_3d_962',['operator=',['../classdlaf_1_1memory_1_1_memory_chunk.html#a0c03f967f09e963519dae5e235da66e1',1,'dlaf::memory::MemoryChunk']]], - ['operator_3d_3d_963',['operator==',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a95e16def9c322ee963d123635c2c875b',1,'dlaf::common::internal::basic_coords::operator==()'],['../classdlaf_1_1common_1_1_size2_d.html#a83385feaa5e14b4291c6267bc68fbe51',1,'dlaf::common::Size2D::operator==()'],['../classdlaf_1_1common_1_1_index2_d.html#a4724fdad093b5d189c5f5e0b1ac5e305',1,'dlaf::common::Index2D::operator==()']]], - ['orthogonal_964',['orthogonal',['../index2d_8h.html#a9b3be798b052f32def2a280a80e2295c',1,'dlaf']]] + ['offset_954',['offset',['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7229ff87452798d32b7195fe684fff7a',1,'dlaf::matrix::internal::View']]], + ['offsetelement_955',['offsetElement',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a398b31442f71b68e39a3ecbb2762203e',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::offsetElement()'],['../structdlaf_1_1matrix_1_1internal_1_1_view.html#a7164cccc46c63cc96661288809a37f5f',1,'dlaf::matrix::internal::View::offsetElement()']]], + ['operator_20bool_956',['operator bool',['../structdlaf_1_1common_1_1_buffer.html#ae57231c34ebf6f79da4d1e6e1c75cf0a',1,'dlaf::common::Buffer::operator bool()'],['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a8f6d65a5adc84620ed33a7de52bac602',1,'dlaf::comm::internal::type_handler::operator bool()']]], + ['operator_20mpi_5fcomm_957',['operator MPI_Comm',['../classdlaf_1_1comm_1_1_communicator.html#a5be809bbc2d038669e31a9de501a2070',1,'dlaf::comm::Communicator']]], + ['operator_20mpi_5fdatatype_958',['operator MPI_Datatype',['../structdlaf_1_1comm_1_1internal_1_1type__handler.html#a2247361b0d7dd77039860c0cd21a13e3',1,'dlaf::comm::internal::type_handler']]], + ['operator_21_3d_959',['operator!=',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#abf2886043fbe929f1e1f0eff5e20935d',1,'dlaf::common::internal::basic_coords::operator!=()'],['../classdlaf_1_1common_1_1_size2_d.html#a1a6822fefe13303612ce0ddf44f260ea',1,'dlaf::common::Size2D::operator!=()'],['../classdlaf_1_1common_1_1_index2_d.html#ae2e28cf6eab4febd97d6c46d23c6096d',1,'dlaf::common::Index2D::operator!=()']]], + ['operator_26_960',['operator&',['../classdlaf_1_1comm_1_1_communicator.html#ace61947783c442bf684c2f035093acf2',1,'dlaf::comm::Communicator']]], + ['operator_28_29_961',['operator()',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#ac22706782f9864d6357df2f48f15ee74',1,'dlaf::matrix::Tile< const T, D >::operator()()'],['../classdlaf_1_1memory_1_1_memory_view.html#aad118548f59d6c8021eb4fbe816a1561',1,'dlaf::memory::MemoryView::operator()() const'],['../classdlaf_1_1memory_1_1_memory_view.html#a6f9c8123e686e2f2122ec988ca0b9725',1,'dlaf::memory::MemoryView::operator()(SizeType index) const'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a1720c41c3c3cbad2e198e7cedbfa84de',1,'dlaf::memory::MemoryChunk::operator()()'],['../classdlaf_1_1memory_1_1_memory_chunk.html#ac0325c028366b219597dde97814e9a7a',1,'dlaf::memory::MemoryChunk::operator()(SizeType index)'],['../classdlaf_1_1matrix_1_1_tile.html#a5641fc26b7eb759a31ddc24975393ce7',1,'dlaf::matrix::Tile::operator()()'],['../classdlaf_1_1common_1_1_pipeline.html#a388b1fe53e84dde5b0f026e7b334902d',1,'dlaf::common::Pipeline::operator()()']]], + ['operator_2b_962',['operator+',['../index2d_8h.html#a2ed7f0c019788f600bc067956646bf75',1,'dlaf::common']]], + ['operator_3c_3c_963',['operator<<',['../index2d_8h.html#a352cdd70886a2a2ace1abd246951c538',1,'dlaf::common']]], + ['operator_3d_964',['operator=',['../classdlaf_1_1memory_1_1_memory_chunk.html#a0c03f967f09e963519dae5e235da66e1',1,'dlaf::memory::MemoryChunk']]], + ['operator_3d_3d_965',['operator==',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a95e16def9c322ee963d123635c2c875b',1,'dlaf::common::internal::basic_coords::operator==()'],['../classdlaf_1_1common_1_1_size2_d.html#a83385feaa5e14b4291c6267bc68fbe51',1,'dlaf::common::Size2D::operator==()'],['../classdlaf_1_1common_1_1_index2_d.html#a4724fdad093b5d189c5f5e0b1ac5e305',1,'dlaf::common::Index2D::operator==()']]], + ['orthogonal_966',['orthogonal',['../index2d_8h.html#a9b3be798b052f32def2a280a80e2295c',1,'dlaf']]] ]; diff --git a/master/search/functions_d.js b/master/search/functions_d.js index 14c654df5d..0d5300b42a 100644 --- a/master/search/functions_d.js +++ b/master/search/functions_d.js @@ -1,11 +1,11 @@ var searchData= [ - ['panel_965',['Panel',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ad63a4d555cea1fdf2f6d5745125fe3cb',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['parentdistribution_966',['parentDistribution',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a66c9a42baab1741e5b562ddac6c716b2',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['permute_967',['permute',['../permutations_2general_8h.html#af301d1e323dbafdc8503756ccd42edcd',1,'dlaf::permutations::permute(SizeType i_begin, SizeType i_end, Matrix< const SizeType, D > &perms, Matrix< const T, D > &mat_in, Matrix< T, D > &mat_out)'],['../permutations_2general_8h.html#a9dbcd748dd76ddf1b2dfef0d08bba03c',1,'dlaf::permutations::permute(comm::CommunicatorGrid grid, common::Pipeline< comm::Communicator > &sub_task_chain, SizeType i_begin, SizeType i_end, Matrix< const SizeType, D > &perms, Matrix< const T, D > &mat_in, Matrix< T, D > &mat_out)']]], - ['pipeline_968',['Pipeline',['../classdlaf_1_1common_1_1_pipeline.html#a6379e14f25ed8d98d3b3434ec0d54a62',1,'dlaf::common::Pipeline']]], - ['potrf_969',['potrf',['../lapack_2tile_8h.html#aab0795227301a9e8e63232028e331fe0',1,'dlaf::tile']]], - ['potrfinfo_970',['potrfInfo',['../lapack_2tile_8h.html#aa917a1efbd53213821776d7d632ee41c',1,'dlaf::tile']]], - ['print_971',['print',['../print__csv_8h.html#ace48c87b6c1d3dc59bf005fb943c0629',1,'dlaf::matrix::print(format::csv, const Tile< const T, Device::CPU > &tile, std::ostream &os=std::cout)'],['../print__csv_8h.html#ae6a4db345c18954cc7d83022a1a848d6',1,'dlaf::matrix::print(format::csv, std::string sym, Matrix< const T, Device::CPU > &mat, std::ostream &os=std::cout)'],['../print__gpu_8h.html#aa0268964c774b355f5513ae93a23b258',1,'dlaf::matrix::print(Format format, const Tile< const T, Device::GPU > &tile, std::ostream &os, whip::stream_t stream)'],['../print__numpy_8h.html#ab467c4a97112f7b1d2cffe3acbb41643',1,'dlaf::matrix::print(format::numpy, const Tile< const T, Device::CPU > &tile, std::ostream &os=std::cout)']]], - ['ptr_972',['ptr',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a018a34c9be26937cbb0c6dbb727c69ce',1,'dlaf::matrix::Tile< const T, D >::ptr() const noexcept'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a7b8f0d5579d0f38743f123046780e860',1,'dlaf::matrix::Tile< const T, D >::ptr(const TileElementIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_tile.html#ad14d471389721941e830f73d7ae0716d',1,'dlaf::matrix::Tile::ptr() const noexcept'],['../classdlaf_1_1matrix_1_1_tile.html#a9d16161523f90ce0f432af6e3eacb9f2',1,'dlaf::matrix::Tile::ptr(const TileElementIndex &index) const noexcept']]] + ['panel_967',['Panel',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ad63a4d555cea1fdf2f6d5745125fe3cb',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['parentdistribution_968',['parentDistribution',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a66c9a42baab1741e5b562ddac6c716b2',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['permute_969',['permute',['../permutations_2general_8h.html#af301d1e323dbafdc8503756ccd42edcd',1,'dlaf::permutations::permute(SizeType i_begin, SizeType i_end, Matrix< const SizeType, D > &perms, Matrix< const T, D > &mat_in, Matrix< T, D > &mat_out)'],['../permutations_2general_8h.html#a9dbcd748dd76ddf1b2dfef0d08bba03c',1,'dlaf::permutations::permute(comm::CommunicatorGrid grid, common::Pipeline< comm::Communicator > &sub_task_chain, SizeType i_begin, SizeType i_end, Matrix< const SizeType, D > &perms, Matrix< const T, D > &mat_in, Matrix< T, D > &mat_out)']]], + ['pipeline_970',['Pipeline',['../classdlaf_1_1common_1_1_pipeline.html#a6379e14f25ed8d98d3b3434ec0d54a62',1,'dlaf::common::Pipeline']]], + ['potrf_971',['potrf',['../lapack_2tile_8h.html#aab0795227301a9e8e63232028e331fe0',1,'dlaf::tile']]], + ['potrfinfo_972',['potrfInfo',['../lapack_2tile_8h.html#aa917a1efbd53213821776d7d632ee41c',1,'dlaf::tile']]], + ['print_973',['print',['../print__csv_8h.html#ace48c87b6c1d3dc59bf005fb943c0629',1,'dlaf::matrix::print(format::csv, const Tile< const T, Device::CPU > &tile, std::ostream &os=std::cout)'],['../print__csv_8h.html#ae6a4db345c18954cc7d83022a1a848d6',1,'dlaf::matrix::print(format::csv, std::string sym, Matrix< const T, Device::CPU > &mat, std::ostream &os=std::cout)'],['../print__gpu_8h.html#aa0268964c774b355f5513ae93a23b258',1,'dlaf::matrix::print(Format format, const Tile< const T, Device::GPU > &tile, std::ostream &os, whip::stream_t stream)'],['../print__numpy_8h.html#ab467c4a97112f7b1d2cffe3acbb41643',1,'dlaf::matrix::print(format::numpy, const Tile< const T, Device::CPU > &tile, std::ostream &os=std::cout)']]], + ['ptr_974',['ptr',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a018a34c9be26937cbb0c6dbb727c69ce',1,'dlaf::matrix::Tile< const T, D >::ptr() const noexcept'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a7b8f0d5579d0f38743f123046780e860',1,'dlaf::matrix::Tile< const T, D >::ptr(const TileElementIndex &index) const noexcept'],['../classdlaf_1_1matrix_1_1_tile.html#ad14d471389721941e830f73d7ae0716d',1,'dlaf::matrix::Tile::ptr() const noexcept'],['../classdlaf_1_1matrix_1_1_tile.html#a9d16161523f90ce0f432af6e3eacb9f2',1,'dlaf::matrix::Tile::ptr(const TileElementIndex &index) const noexcept']]] ]; diff --git a/master/search/functions_e.js b/master/search/functions_e.js index 6d834493b7..af9e3591e9 100644 --- a/master/search/functions_e.js +++ b/master/search/functions_e.js @@ -1,25 +1,26 @@ var searchData= [ - ['rangeend_973',['rangeEnd',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a3aa72e4de5ae6c2d87b0643e08311df8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['rangeendlocal_974',['rangeEndLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a7aa91d49db2332c93e5b2b2e1aadbecf',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['rangestart_975',['rangeStart',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a09572f6a6461327f0ed6d0e471174410',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['rangestartlocal_976',['rangeStartLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#acef133d14867a04758e8914135c5d5f8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['rank_977',['rank',['../classdlaf_1_1comm_1_1_communicator.html#a52be17b02755dcbfbdcaa95ba5f9230c',1,'dlaf::comm::Communicator::rank()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#a45574d05dd3cfb3d40d198cdf10b169e',1,'dlaf::comm::CommunicatorGrid::rank() const noexcept']]], - ['rankfullcommunicator_978',['rankFullCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#ad9a0a0a05723adbe0e57171c7c1e2e93',1,'dlaf::comm::CommunicatorGrid']]], - ['rankglobalelement_979',['rankGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a8eac06b3be1805d7261a75809529a910',1,'dlaf::matrix::Distribution']]], - ['rankglobaltile_980',['rankGlobalTile',['../classdlaf_1_1matrix_1_1_distribution.html#ae3f7cda513d213773cb4d8a0ba22b758',1,'dlaf::matrix::Distribution::rankGlobalTile(const GlobalTileIndex &global_tile) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#a49ff769692620e7f94cf796a65ff6a9e',1,'dlaf::matrix::Distribution::rankGlobalTile(SizeType global_tile) const noexcept'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a622f4764a792986112c89f282e04080f',1,'dlaf::matrix::internal::MatrixBase::rankGlobalTile()'],['../util__distribution_8h.html#ad89b3802ea0ed39625419b53e3f199d1',1,'dlaf::util::matrix::rankGlobalTile()']]], - ['rankindex_981',['rankIndex',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a685f321909fc853e023bff712c964c2d',1,'dlaf::matrix::internal::MatrixBase']]], - ['read_982',['read',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#abd8114d7c9ff03a2ff5c8613284135f6',1,'dlaf::matrix::Matrix< const T, D >::read(const LocalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#a8485d802a7f318a845de46e1b4e48118',1,'dlaf::matrix::Matrix< const T, D >::read(const GlobalTileIndex &index)'],['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a4631ad4a6fa72e74cf95798c527a48d1',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::read()'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#a814972188382d0ae0120b6acfddd92bb',1,'dlaf::matrix::RetiledMatrix::read(const LocalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#afacf9af5a45a341c284844d866227579',1,'dlaf::matrix::RetiledMatrix::read(const GlobalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a99845f6d6b238d3a8d543d20d3adb04d',1,'dlaf::matrix::internal::TilePipeline::read()']]], - ['readwrite_983',['readwrite',['../classdlaf_1_1matrix_1_1_retiled_matrix.html#ac3ca2e52e6627f5890bcb25fcff60167',1,'dlaf::matrix::RetiledMatrix::readwrite()'],['../classdlaf_1_1matrix_1_1_matrix.html#af486e66243b57df94b9822bd2553ae38',1,'dlaf::matrix::Matrix::readwrite()'],['../classdlaf_1_1matrix_1_1_retiled_matrix.html#aa0e45d9a97f8d3ff418d2bd08211a5bf',1,'dlaf::matrix::RetiledMatrix::readwrite()'],['../classdlaf_1_1matrix_1_1_matrix.html#ad11e9036fe317b6328ae71585b19c811',1,'dlaf::matrix::Matrix::readwrite()'],['../structdlaf_1_1matrix_1_1_panel.html#aac41b4dfa833e4a24839fe1d470b835a',1,'dlaf::matrix::Panel::readwrite()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#ab0a26ee4322ab4b7a3d250a82d072f68',1,'dlaf::matrix::internal::TilePipeline::readwrite()']]], - ['readwrite_5fwith_5fwrapper_984',['readwrite_with_wrapper',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a17403152214a88994a75395f7bebfe3c',1,'dlaf::matrix::internal::TilePipeline']]], - ['receive_5ffrom_985',['receive_from',['../basic_8h.html#a13040096b22855cefca83ec02f0c64c0',1,'dlaf::comm::sync::receive_from()'],['../sync_2broadcast_8h.html#a86fb5d23e8299cbd380a4e7d724c1112',1,'dlaf::comm::sync::broadcast::receive_from()']]], - ['reduce_986',['reduce',['../sync_2reduce_8h.html#a22dfe268f1dcd2f08e2678af288063be',1,'dlaf::comm::sync']]], - ['reduceinplace_987',['reduceInPlace',['../sync_2reduce_8h.html#a86b20000f6542c605c6b5f2bcc9e625b',1,'dlaf::comm::sync']]], - ['reducerecv_988',['reduceRecv',['../sync_2reduce_8h.html#a03e9b0551a974dfa12fc9885272a1bee',1,'dlaf::comm::sync']]], - ['reducerecvinplace_989',['reduceRecvInPlace',['../sync_2reduce_8h.html#a1ce87e8bcfccf05e7edd9d23e1f778f3',1,'dlaf::comm::sync']]], - ['reducesend_990',['reduceSend',['../sync_2reduce_8h.html#a27df5989f95087d0259a3e381b1539b3',1,'dlaf::comm::sync']]], - ['reductiontoband_991',['reductionToBand',['../reduction__to__band_8h.html#a00574f61da45042c5276b043d7c33617',1,'dlaf::eigensolver::reductionToBand(Matrix< T, D > &mat_a, const SizeType band_size)'],['../reduction__to__band_8h.html#a143b74447a1f72d4d5368b9bad10b331',1,'dlaf::eigensolver::reductionToBand(comm::CommunicatorGrid grid, Matrix< T, D > &mat_a, const SizeType band_size)']]], - ['reset_992',['reset',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#acb87326419c25da556f027565b96c086',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::reset()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a83ffbae7e4944baa1bc4bfbe5dd1100d',1,'dlaf::matrix::internal::TilePipeline::reset()'],['../classdlaf_1_1common_1_1_pipeline.html#a4a9e1a35ffad351b5b4e01b88ef8a447',1,'dlaf::common::Pipeline::reset()']]], - ['retiledmatrix_993',['RetiledMatrix',['../classdlaf_1_1matrix_1_1_retiled_matrix.html#a4e208b9159a478b3fa20167ec8e04318',1,'dlaf::matrix::RetiledMatrix']]], - ['rowcommunicator_994',['rowCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#ab13487adb2992d2cc50e06b498582391',1,'dlaf::comm::CommunicatorGrid']]] + ['rangeend_975',['rangeEnd',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a3aa72e4de5ae6c2d87b0643e08311df8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['rangeendlocal_976',['rangeEndLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a7aa91d49db2332c93e5b2b2e1aadbecf',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['rangestart_977',['rangeStart',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a09572f6a6461327f0ed6d0e471174410',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['rangestartlocal_978',['rangeStartLocal',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#acef133d14867a04758e8914135c5d5f8',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['rank_979',['rank',['../classdlaf_1_1comm_1_1_communicator.html#a52be17b02755dcbfbdcaa95ba5f9230c',1,'dlaf::comm::Communicator::rank()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#a45574d05dd3cfb3d40d198cdf10b169e',1,'dlaf::comm::CommunicatorGrid::rank() const noexcept']]], + ['rankfullcommunicator_980',['rankFullCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#ad9a0a0a05723adbe0e57171c7c1e2e93',1,'dlaf::comm::CommunicatorGrid']]], + ['rankglobalelement_981',['rankGlobalElement',['../classdlaf_1_1matrix_1_1_distribution.html#a8eac06b3be1805d7261a75809529a910',1,'dlaf::matrix::Distribution']]], + ['rankglobaltile_982',['rankGlobalTile',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a622f4764a792986112c89f282e04080f',1,'dlaf::matrix::internal::MatrixBase::rankGlobalTile()'],['../util__distribution_8h.html#ad89b3802ea0ed39625419b53e3f199d1',1,'dlaf::util::matrix::rankGlobalTile()'],['../classdlaf_1_1matrix_1_1_distribution.html#a49ff769692620e7f94cf796a65ff6a9e',1,'dlaf::matrix::Distribution::rankGlobalTile(SizeType global_tile) const noexcept'],['../classdlaf_1_1matrix_1_1_distribution.html#ae3f7cda513d213773cb4d8a0ba22b758',1,'dlaf::matrix::Distribution::rankGlobalTile(const GlobalTileIndex &global_tile) const noexcept']]], + ['rankindex_983',['rankIndex',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a685f321909fc853e023bff712c964c2d',1,'dlaf::matrix::internal::MatrixBase']]], + ['read_984',['read',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a99845f6d6b238d3a8d543d20d3adb04d',1,'dlaf::matrix::internal::TilePipeline::read()'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#abd8114d7c9ff03a2ff5c8613284135f6',1,'dlaf::matrix::Matrix< const T, D >::read(const LocalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#a8485d802a7f318a845de46e1b4e48118',1,'dlaf::matrix::Matrix< const T, D >::read(const GlobalTileIndex &index)'],['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a4631ad4a6fa72e74cf95798c527a48d1',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::read()']]], + ['readwrite_985',['readwrite',['../structdlaf_1_1matrix_1_1_panel.html#aac41b4dfa833e4a24839fe1d470b835a',1,'dlaf::matrix::Panel::readwrite()'],['../classdlaf_1_1matrix_1_1_matrix.html#af486e66243b57df94b9822bd2553ae38',1,'dlaf::matrix::Matrix::readwrite(const GlobalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1_matrix.html#ad11e9036fe317b6328ae71585b19c811',1,'dlaf::matrix::Matrix::readwrite(const LocalTileIndex &index) noexcept'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#ab0a26ee4322ab4b7a3d250a82d072f68',1,'dlaf::matrix::internal::TilePipeline::readwrite()']]], + ['readwrite_5fwith_5fwrapper_986',['readwrite_with_wrapper',['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a17403152214a88994a75395f7bebfe3c',1,'dlaf::matrix::internal::TilePipeline']]], + ['receive_5ffrom_987',['receive_from',['../basic_8h.html#a13040096b22855cefca83ec02f0c64c0',1,'dlaf::comm::sync::receive_from()'],['../sync_2broadcast_8h.html#a86fb5d23e8299cbd380a4e7d724c1112',1,'dlaf::comm::sync::broadcast::receive_from()']]], + ['reduce_988',['reduce',['../sync_2reduce_8h.html#a22dfe268f1dcd2f08e2678af288063be',1,'dlaf::comm::sync']]], + ['reduceinplace_989',['reduceInPlace',['../sync_2reduce_8h.html#a86b20000f6542c605c6b5f2bcc9e625b',1,'dlaf::comm::sync']]], + ['reducerecv_990',['reduceRecv',['../sync_2reduce_8h.html#a03e9b0551a974dfa12fc9885272a1bee',1,'dlaf::comm::sync']]], + ['reducerecvinplace_991',['reduceRecvInPlace',['../sync_2reduce_8h.html#a1ce87e8bcfccf05e7edd9d23e1f778f3',1,'dlaf::comm::sync']]], + ['reducesend_992',['reduceSend',['../sync_2reduce_8h.html#a27df5989f95087d0259a3e381b1539b3',1,'dlaf::comm::sync']]], + ['reductiontoband_993',['reductionToBand',['../reduction__to__band_8h.html#a00574f61da45042c5276b043d7c33617',1,'dlaf::eigensolver::reductionToBand(Matrix< T, D > &mat_a, const SizeType band_size)'],['../reduction__to__band_8h.html#a143b74447a1f72d4d5368b9bad10b331',1,'dlaf::eigensolver::reductionToBand(comm::CommunicatorGrid grid, Matrix< T, D > &mat_a, const SizeType band_size)']]], + ['reset_994',['reset',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#acb87326419c25da556f027565b96c086',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::reset()'],['../classdlaf_1_1matrix_1_1internal_1_1_tile_pipeline.html#a83ffbae7e4944baa1bc4bfbe5dd1100d',1,'dlaf::matrix::internal::TilePipeline::reset()'],['../classdlaf_1_1common_1_1_pipeline.html#a4a9e1a35ffad351b5b4e01b88ef8a447',1,'dlaf::common::Pipeline::reset()']]], + ['retiledsubpipeline_995',['retiledSubPipeline',['../classdlaf_1_1matrix_1_1_matrix.html#aae1c8b22f9862944a006003f528595ff',1,'dlaf::matrix::Matrix']]], + ['retiledsubpipelineconst_996',['retiledSubPipelineConst',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#aa10512ea4bde924a55e73bfe07050942',1,'dlaf::matrix::Matrix< const T, D >']]], + ['rowcommunicator_997',['rowCommunicator',['../classdlaf_1_1comm_1_1_communicator_grid.html#ab13487adb2992d2cc50e06b498582391',1,'dlaf::comm::CommunicatorGrid']]] ]; diff --git a/master/search/functions_f.js b/master/search/functions_f.js index 561677613c..e12312f80e 100644 --- a/master/search/functions_f.js +++ b/master/search/functions_f.js @@ -1,41 +1,44 @@ var searchData= [ - ['samesign_995',['sameSign',['../util__math_8h.html#a0922553e7cd5d09cc857bb33e8a1b91b',1,'dlaf::util']]], - ['scheduleallreduce_996',['scheduleAllReduce',['../kernels_2all__reduce_8h.html#a8cb64ca1859aef0b95cdcfa8a1f05297',1,'dlaf::comm']]], - ['scheduleallreduceinplace_997',['scheduleAllReduceInPlace',['../kernels_2all__reduce_8h.html#a689b3d06eea6e5d2c9899521e970ecb0',1,'dlaf::comm']]], - ['scheduleallsump2p_998',['scheduleAllSumP2P',['../p2p__allsum_8h.html#ac786ec776d72b61529c8a79c5baffff4',1,'dlaf::comm']]], - ['schedulerecvbcast_999',['scheduleRecvBcast',['../kernels_2broadcast_8h.html#a3645d0cfc3caad43c37c355896a7215f',1,'dlaf::comm']]], - ['schedulereducerecvinplace_1000',['scheduleReduceRecvInPlace',['../kernels_2reduce_8h.html#a89af8ab199d6ba8902cc962122477f00',1,'dlaf::comm']]], - ['schedulereducesend_1001',['scheduleReduceSend',['../kernels_2reduce_8h.html#a60899e3a349b0defe79b1561045a373a',1,'dlaf::comm']]], - ['schedulesendbcast_1002',['scheduleSendBcast',['../kernels_2broadcast_8h.html#a98b583f24caaccdf37f3db1f3b46b457',1,'dlaf::comm']]], - ['select_1003',['select',['../matrix_8h.html#a1abc47c06dc1c46c9bdc2425090a02ab',1,'dlaf::matrix']]], - ['selectgeneric_1004',['selectGeneric',['../matrix_8h.html#ac3737882aaaf23b1f901da6396202ba1',1,'dlaf::matrix::internal']]], - ['selectread_1005',['selectRead',['../matrix_8h.html#a02db5399b6bc7e53a162f5939ed7ca20',1,'dlaf::matrix']]], - ['send_1006',['send',['../sync_2broadcast_8h.html#a313a9dc3d212f70d7af273fa0c681aee',1,'dlaf::comm::sync::broadcast']]], - ['send_5fto_1007',['send_to',['../basic_8h.html#adfd3205605f800afb2208146565a3bd4',1,'dlaf::comm::sync']]], - ['set_1008',['set',['../util__matrix_8h.html#a1f796f1fd4dd705c90ac141d687843b2',1,'dlaf::matrix::util::set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f)'],['../util__matrix_8h.html#af07478cd4a633a1422e160863d3d8510',1,'dlaf::matrix::util::set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f, const blas::Op op)']]], - ['set0_1009',['set0',['../lapack_2tile_8h.html#ab094eeac6052422b50a298b42a947251',1,'dlaf::tile::set0()'],['../util__matrix_8h.html#a6cdf6afc18a6456ed68b3fd1dffa9c39',1,'dlaf::matrix::util::set0(pika::execution::thread_priority priority, LocalTileIndex begin, LocalTileSize sz, Matrix< T, D > &matrix)'],['../util__matrix_8h.html#a79688b3669b67c143114cf2eef3ba69b',1,'dlaf::matrix::util::set0(pika::execution::thread_priority priority, Panel< axis, T, D, storage > &panel)']]], - ['set_5frandom_1010',['set_random',['../util__matrix_8h.html#a8e8c0d0551fbda5968d6c84eb3c70fc0',1,'dlaf::matrix::util']]], - ['set_5frandom_5fhermitian_1011',['set_random_hermitian',['../util__matrix_8h.html#a3f7866203ad62c615d11ca18af4d4476',1,'dlaf::matrix::util']]], - ['set_5frandom_5fhermitian_5fpositive_5fdefinite_1012',['set_random_hermitian_positive_definite',['../util__matrix_8h.html#ae2afbf7cdec8225e544b7b71aa52ee58',1,'dlaf::matrix::util']]], - ['set_5frandom_5fhermitian_5fwith_5foffset_1013',['set_random_hermitian_with_offset',['../util__matrix_8h.html#a2f1823d0004b53105c4d4b7373208133',1,'dlaf::matrix::util::internal']]], - ['setheight_1014',['setHeight',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a5ffd11dcc29ca2731c0cf232cb599d8e',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setrange_1015',['setRange',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a88f387bed0a5cf8dabd5f6c5b3d2c3b0',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setrangeend_1016',['setRangeEnd',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a542319871ee2379e84651d7dd91f0749',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setrangestart_1017',['setRangeStart',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a621d6ffd26cb962792c4087159df2769',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['settile_1018',['setTile',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa9c0ce49a7829488736d99011c5ac6c9',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setupinternalmatrix_1019',['setupInternalMatrix',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a11f5026c35f76adb7c99c725dc969c44',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['setwidth_1020',['setWidth',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a74e5d5d87edfb1d2de8a54cfa812bd04',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['sharereadwritetile_1021',['shareReadWriteTile',['../matrix_2tile_8h.html#aa8e4a64946c725cef06834fb80b75276',1,'dlaf::matrix']]], - ['size_1022',['size',['../classdlaf_1_1memory_1_1_memory_view.html#a5ae9491601035e4f2fd87352916f84c2',1,'dlaf::memory::MemoryView::size()'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a886c87812f0a8eb9071c70477ac9f042',1,'dlaf::memory::MemoryChunk::size()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a7ef713ad7d24caccfc0863ab4b8481f2',1,'dlaf::matrix::Tile< const T, D >::size()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#adc5a25970dab7cdea07e2f98328df805',1,'dlaf::matrix::internal::MatrixBase::size()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#a9307e8d79bffad58211c317c92c7dc77',1,'dlaf::comm::CommunicatorGrid::size()'],['../classdlaf_1_1comm_1_1_communicator.html#a415f5d36c3033320074df329163c3127',1,'dlaf::comm::Communicator::size()']]], - ['splittile_1023',['splitTile',['../matrix_2tile_8h.html#a7a2383f364417fce38d6e7038cb18bd1',1,'dlaf::matrix::splitTile(ReadOnlyTileSender< T, D > tile, const std::vector< SubTileSpec > &specs)'],['../matrix_2tile_8h.html#ae46de76ccc35985387f6820b1db9fb51',1,'dlaf::matrix::splitTile(ReadOnlyTileSender< T, D > tile, const SubTileSpec &spec)'],['../matrix_2tile_8h.html#ae5697bd5c424a95890978dfdc4739bf6',1,'dlaf::matrix::splitTile(ReadWriteTileSender< T, D > &&tile, const SubTileSpec &spec)']]], - ['splittiledisjoint_1024',['splitTileDisjoint',['../matrix_2tile_8h.html#a4c15e2e57d6336347e6e19d945bd650e',1,'dlaf::matrix']]], - ['square_5fblocksize_1025',['square_blocksize',['../util__matrix_8h.html#a355682eaf279129f2aa06c8e96eaa61b',1,'dlaf::matrix']]], - ['square_5fsize_1026',['square_size',['../util__matrix_8h.html#aae6464887dae32e01668a32e71397cb0',1,'dlaf::matrix::square_size()'],['../util__tile_8h.html#adfd1e0b37d590f9f6b56cdb276688f0f',1,'dlaf::tile::square_size(const Tile< T, D > &t) noexcept']]], - ['stedc_1027',['stedc',['../lapack_2tile_8h.html#a68163899b7b15f447fdbb05e475d9cc6',1,'dlaf::tile']]], - ['stride_1028',['stride',['../structdlaf_1_1common_1_1_data_descriptor.html#ab4c4694a381e4cfb44d304cf150743f9',1,'dlaf::common::DataDescriptor']]], - ['submatrixview_1029',['SubMatrixView',['../structdlaf_1_1matrix_1_1_sub_matrix_view.html#a7c7cde76a018f82ab23c749de90fe46d',1,'dlaf::matrix::SubMatrixView']]], - ['subpanelview_1030',['SubPanelView',['../structdlaf_1_1matrix_1_1_sub_panel_view.html#a556b70975f5928ca81f2196d549be232',1,'dlaf::matrix::SubPanelView']]], - ['subtilereference_1031',['subTileReference',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a4adf525073d06baaa47b06886272da41',1,'dlaf::matrix::Tile< const T, D >::subTileReference()'],['../classdlaf_1_1matrix_1_1_tile.html#a670161e49f6a18b5979e9938e4e7e867',1,'dlaf::matrix::Tile::subTileReference()']]], - ['sum_1032',['sum',['../util__math_8h.html#afa826e9affbc0fc9e9dbe24d233a9ffb',1,'dlaf::util::size_t::sum()'],['../util__math_8h.html#ae1dbf257cc807e5bdba5a63087e3a2bc',1,'dlaf::util::ptrdiff_t::sum()']]] + ['samesign_998',['sameSign',['../util__math_8h.html#a0922553e7cd5d09cc857bb33e8a1b91b',1,'dlaf::util']]], + ['scheduleallreduce_999',['scheduleAllReduce',['../kernels_2all__reduce_8h.html#a8cb64ca1859aef0b95cdcfa8a1f05297',1,'dlaf::comm']]], + ['scheduleallreduceinplace_1000',['scheduleAllReduceInPlace',['../kernels_2all__reduce_8h.html#a689b3d06eea6e5d2c9899521e970ecb0',1,'dlaf::comm']]], + ['scheduleallsump2p_1001',['scheduleAllSumP2P',['../p2p__allsum_8h.html#ac786ec776d72b61529c8a79c5baffff4',1,'dlaf::comm']]], + ['schedulerecvbcast_1002',['scheduleRecvBcast',['../kernels_2broadcast_8h.html#a3645d0cfc3caad43c37c355896a7215f',1,'dlaf::comm']]], + ['schedulereducerecvinplace_1003',['scheduleReduceRecvInPlace',['../kernels_2reduce_8h.html#a89af8ab199d6ba8902cc962122477f00',1,'dlaf::comm']]], + ['schedulereducesend_1004',['scheduleReduceSend',['../kernels_2reduce_8h.html#a60899e3a349b0defe79b1561045a373a',1,'dlaf::comm']]], + ['schedulesendbcast_1005',['scheduleSendBcast',['../kernels_2broadcast_8h.html#a98b583f24caaccdf37f3db1f3b46b457',1,'dlaf::comm']]], + ['select_1006',['select',['../matrix_8h.html#a1abc47c06dc1c46c9bdc2425090a02ab',1,'dlaf::matrix']]], + ['selectgeneric_1007',['selectGeneric',['../matrix_8h.html#ac3737882aaaf23b1f901da6396202ba1',1,'dlaf::matrix::internal']]], + ['selectread_1008',['selectRead',['../matrix_8h.html#a02db5399b6bc7e53a162f5939ed7ca20',1,'dlaf::matrix']]], + ['send_1009',['send',['../sync_2broadcast_8h.html#a313a9dc3d212f70d7af273fa0c681aee',1,'dlaf::comm::sync::broadcast']]], + ['send_5fto_1010',['send_to',['../basic_8h.html#adfd3205605f800afb2208146565a3bd4',1,'dlaf::comm::sync']]], + ['set_1011',['set',['../util__matrix_8h.html#a1f796f1fd4dd705c90ac141d687843b2',1,'dlaf::matrix::util::set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f)'],['../util__matrix_8h.html#af07478cd4a633a1422e160863d3d8510',1,'dlaf::matrix::util::set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f, const blas::Op op)']]], + ['set0_1012',['set0',['../lapack_2tile_8h.html#ab094eeac6052422b50a298b42a947251',1,'dlaf::tile::set0()'],['../util__matrix_8h.html#a6cdf6afc18a6456ed68b3fd1dffa9c39',1,'dlaf::matrix::util::set0(pika::execution::thread_priority priority, LocalTileIndex begin, LocalTileSize sz, Matrix< T, D > &matrix)'],['../util__matrix_8h.html#a79688b3669b67c143114cf2eef3ba69b',1,'dlaf::matrix::util::set0(pika::execution::thread_priority priority, Panel< axis, T, D, storage > &panel)']]], + ['set_5frandom_1013',['set_random',['../util__matrix_8h.html#a8e8c0d0551fbda5968d6c84eb3c70fc0',1,'dlaf::matrix::util']]], + ['set_5frandom_5fhermitian_1014',['set_random_hermitian',['../util__matrix_8h.html#a3f7866203ad62c615d11ca18af4d4476',1,'dlaf::matrix::util']]], + ['set_5frandom_5fhermitian_5fpositive_5fdefinite_1015',['set_random_hermitian_positive_definite',['../util__matrix_8h.html#ae2afbf7cdec8225e544b7b71aa52ee58',1,'dlaf::matrix::util']]], + ['set_5frandom_5fhermitian_5fwith_5foffset_1016',['set_random_hermitian_with_offset',['../util__matrix_8h.html#a2f1823d0004b53105c4d4b7373208133',1,'dlaf::matrix::util::internal']]], + ['setheight_1017',['setHeight',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a5ffd11dcc29ca2731c0cf232cb599d8e',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setrange_1018',['setRange',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a88f387bed0a5cf8dabd5f6c5b3d2c3b0',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setrangeend_1019',['setRangeEnd',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a542319871ee2379e84651d7dd91f0749',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setrangestart_1020',['setRangeStart',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a621d6ffd26cb962792c4087159df2769',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['settile_1021',['setTile',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#aa9c0ce49a7829488736d99011c5ac6c9',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setupinternalmatrix_1022',['setupInternalMatrix',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a11f5026c35f76adb7c99c725dc969c44',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['setwidth_1023',['setWidth',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a74e5d5d87edfb1d2de8a54cfa812bd04',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['sharereadwritetile_1024',['shareReadWriteTile',['../matrix_2tile_8h.html#aa8e4a64946c725cef06834fb80b75276',1,'dlaf::matrix']]], + ['single_5ftile_5fper_5fblock_1025',['single_tile_per_block',['../util__matrix_8h.html#a4e47364118c7bc9bb3d038ae4ec8a35c',1,'dlaf::matrix']]], + ['size_1026',['size',['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#adc5a25970dab7cdea07e2f98328df805',1,'dlaf::matrix::internal::MatrixBase::size()'],['../classdlaf_1_1comm_1_1_communicator.html#a415f5d36c3033320074df329163c3127',1,'dlaf::comm::Communicator::size()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#a9307e8d79bffad58211c317c92c7dc77',1,'dlaf::comm::CommunicatorGrid::size()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a7ef713ad7d24caccfc0863ab4b8481f2',1,'dlaf::matrix::Tile< const T, D >::size()'],['../classdlaf_1_1memory_1_1_memory_chunk.html#a886c87812f0a8eb9071c70477ac9f042',1,'dlaf::memory::MemoryChunk::size()'],['../classdlaf_1_1memory_1_1_memory_view.html#a5ae9491601035e4f2fd87352916f84c2',1,'dlaf::memory::MemoryView::size()']]], + ['splittile_1027',['splitTile',['../matrix_2tile_8h.html#ae46de76ccc35985387f6820b1db9fb51',1,'dlaf::matrix::splitTile(ReadOnlyTileSender< T, D > tile, const SubTileSpec &spec)'],['../matrix_2tile_8h.html#a7a2383f364417fce38d6e7038cb18bd1',1,'dlaf::matrix::splitTile(ReadOnlyTileSender< T, D > tile, const std::vector< SubTileSpec > &specs)'],['../matrix_2tile_8h.html#ae5697bd5c424a95890978dfdc4739bf6',1,'dlaf::matrix::splitTile(ReadWriteTileSender< T, D > &&tile, const SubTileSpec &spec)']]], + ['splittiledisjoint_1028',['splitTileDisjoint',['../matrix_2tile_8h.html#a4c15e2e57d6336347e6e19d945bd650e',1,'dlaf::matrix']]], + ['square_5fblocksize_1029',['square_blocksize',['../util__matrix_8h.html#a355682eaf279129f2aa06c8e96eaa61b',1,'dlaf::matrix']]], + ['square_5fsize_1030',['square_size',['../util__tile_8h.html#adfd1e0b37d590f9f6b56cdb276688f0f',1,'dlaf::tile::square_size()'],['../util__matrix_8h.html#aae6464887dae32e01668a32e71397cb0',1,'dlaf::matrix::square_size()']]], + ['stedc_1031',['stedc',['../lapack_2tile_8h.html#a68163899b7b15f447fdbb05e475d9cc6',1,'dlaf::tile']]], + ['stride_1032',['stride',['../structdlaf_1_1common_1_1_data_descriptor.html#ab4c4694a381e4cfb44d304cf150743f9',1,'dlaf::common::DataDescriptor']]], + ['submatrixview_1033',['SubMatrixView',['../structdlaf_1_1matrix_1_1_sub_matrix_view.html#a7c7cde76a018f82ab23c749de90fe46d',1,'dlaf::matrix::SubMatrixView']]], + ['subpanelview_1034',['SubPanelView',['../structdlaf_1_1matrix_1_1_sub_panel_view.html#a556b70975f5928ca81f2196d549be232',1,'dlaf::matrix::SubPanelView']]], + ['subpipeline_1035',['subPipeline',['../classdlaf_1_1matrix_1_1_matrix.html#a720bcfeb6db4c2f2d6e144f8570b8e23',1,'dlaf::matrix::Matrix']]], + ['subpipelineconst_1036',['subPipelineConst',['../classdlaf_1_1matrix_1_1_matrix_3_01const_01_t_00_01_d_01_4.html#aa00bb4dd8e50ec14ad04c8d600982d8c',1,'dlaf::matrix::Matrix< const T, D >']]], + ['subtilereference_1037',['subTileReference',['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a4adf525073d06baaa47b06886272da41',1,'dlaf::matrix::Tile< const T, D >::subTileReference()'],['../classdlaf_1_1matrix_1_1_tile.html#a670161e49f6a18b5979e9938e4e7e867',1,'dlaf::matrix::Tile::subTileReference()']]], + ['sum_1038',['sum',['../util__math_8h.html#afa826e9affbc0fc9e9dbe24d233a9ffb',1,'dlaf::util::size_t::sum()'],['../util__math_8h.html#ae1dbf257cc807e5bdba5a63087e3a2bc',1,'dlaf::util::ptrdiff_t::sum()']]] ]; diff --git a/master/search/related_0.js b/master/search/related_0.js index 55b3cb6422..d8baecd6ba 100644 --- a/master/search/related_0.js +++ b/master/search/related_0.js @@ -1,4 +1,4 @@ var searchData= [ - ['isindexinrange_1085',['isIndexInRange',['../classdlaf_1_1common_1_1_iterable_range2_d.html#a833f81c7ed8c08d0e995d9c97e8aa935',1,'dlaf::common::IterableRange2D']]] + ['isindexinrange_1091',['isIndexInRange',['../classdlaf_1_1common_1_1_iterable_range2_d.html#a833f81c7ed8c08d0e995d9c97e8aa935',1,'dlaf::common::IterableRange2D']]] ]; diff --git a/master/search/related_1.js b/master/search/related_1.js index cdb3f93fcd..cb50fff083 100644 --- a/master/search/related_1.js +++ b/master/search/related_1.js @@ -1,4 +1,4 @@ var searchData= [ - ['make_5fcommunicator_5fmanaged_1086',['make_communicator_managed',['../classdlaf_1_1comm_1_1_communicator.html#ada042e7529d465519d9fb98e004d6a55',1,'dlaf::comm::Communicator']]] + ['make_5fcommunicator_5fmanaged_1092',['make_communicator_managed',['../classdlaf_1_1comm_1_1_communicator.html#ada042e7529d465519d9fb98e004d6a55',1,'dlaf::comm::Communicator']]] ]; diff --git a/master/search/related_2.js b/master/search/related_2.js index 7b42519dd1..3769b7d00a 100644 --- a/master/search/related_2.js +++ b/master/search/related_2.js @@ -1,4 +1,4 @@ var searchData= [ - ['operator_3c_3c_1087',['operator<<',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a9a65958ab367757d900bd588a405df77',1,'dlaf::common::internal::basic_coords::operator<<()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#ac215ee8142401d06aba72494c8aa8b9f',1,'dlaf::comm::CommunicatorGrid::operator<<()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a054e897882d7ab913c1b63af96e87fec',1,'dlaf::matrix::internal::MatrixBase::operator<<()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a45b5556596386e6bb46e0dc61b838811',1,'dlaf::matrix::Tile< const T, D >::operator<<()']]] + ['operator_3c_3c_1093',['operator<<',['../classdlaf_1_1common_1_1internal_1_1basic__coords.html#a9a65958ab367757d900bd588a405df77',1,'dlaf::common::internal::basic_coords::operator<<()'],['../classdlaf_1_1comm_1_1_communicator_grid.html#ac215ee8142401d06aba72494c8aa8b9f',1,'dlaf::comm::CommunicatorGrid::operator<<()'],['../classdlaf_1_1matrix_1_1internal_1_1_matrix_base.html#a054e897882d7ab913c1b63af96e87fec',1,'dlaf::matrix::internal::MatrixBase::operator<<()'],['../classdlaf_1_1matrix_1_1_tile_3_01const_01_t_00_01_d_01_4.html#a45b5556596386e6bb46e0dc61b838811',1,'dlaf::matrix::Tile< const T, D >::operator<<()']]] ]; diff --git a/master/search/typedefs_0.js b/master/search/typedefs_0.js index 5a343d8336..b22fa9d98c 100644 --- a/master/search/typedefs_0.js +++ b/master/search/typedefs_0.js @@ -1,5 +1,5 @@ var searchData= [ - ['enable_5fif_5fconvertible_5ft_1079',['enable_if_convertible_t',['../traits_8h.html#a234bd5c290fe869e9856e7e2d3667baf',1,'dlaf']]], - ['enable_5fif_5fsignature_5ft_1080',['enable_if_signature_t',['../traits_8h.html#ada84be4e014ed8f3d91175b3856104e5',1,'dlaf']]] + ['enable_5fif_5fconvertible_5ft_1085',['enable_if_convertible_t',['../traits_8h.html#a234bd5c290fe869e9856e7e2d3667baf',1,'dlaf']]], + ['enable_5fif_5fsignature_5ft_1086',['enable_if_signature_t',['../traits_8h.html#ada84be4e014ed8f3d91175b3856104e5',1,'dlaf']]] ]; diff --git a/master/search/typedefs_1.js b/master/search/typedefs_1.js index 6c9107e11f..22163d86e2 100644 --- a/master/search/typedefs_1.js +++ b/master/search/typedefs_1.js @@ -1,5 +1,5 @@ var searchData= [ - ['index2d_1081',['Index2D',['../communicator__grid_8h.html#af5204c68f1436b89ad06c653b710432e',1,'dlaf::comm']]], - ['indext_5fmpi_1082',['IndexT_MPI',['../communicator_8h.html#a2b48f62607e5e128dff19ceac22093e9',1,'dlaf::comm']]] + ['index2d_1087',['Index2D',['../communicator__grid_8h.html#af5204c68f1436b89ad06c653b710432e',1,'dlaf::comm']]], + ['indext_5fmpi_1088',['IndexT_MPI',['../communicator_8h.html#a2b48f62607e5e128dff19ceac22093e9',1,'dlaf::comm']]] ]; diff --git a/master/search/typedefs_2.js b/master/search/typedefs_2.js index 6f73d8be08..06de637042 100644 --- a/master/search/typedefs_2.js +++ b/master/search/typedefs_2.js @@ -1,4 +1,4 @@ var searchData= [ - ['size2d_1083',['Size2D',['../communicator__grid_8h.html#a1af44f2ea119a10f873e64cf048ba94a',1,'dlaf::comm']]] + ['size2d_1089',['Size2D',['../communicator__grid_8h.html#a1af44f2ea119a10f873e64cf048ba94a',1,'dlaf::comm']]] ]; diff --git a/master/search/variables_0.js b/master/search/variables_0.js index 4975fba442..e8832c616d 100644 --- a/master/search/variables_0.js +++ b/master/search/variables_0.js @@ -1,4 +1,4 @@ var searchData= [ - ['bias_5f_1068',['bias_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a2915562160f1b12d3f9e0f49b339988a',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] + ['bias_5f_1074',['bias_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a2915562160f1b12d3f9e0f49b339988a',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] ]; diff --git a/master/search/variables_1.js b/master/search/variables_1.js index 0f89c7722d..688601c50e 100644 --- a/master/search/variables_1.js +++ b/master/search/variables_1.js @@ -1,5 +1,5 @@ var searchData= [ - ['classic_5ftype_5f_1069',['classic_type_',['../classdlaf_1_1comm_1_1_message.html#a49412018d3ffdfdbd87690f59fb54a14',1,'dlaf::comm::Message']]], - ['custom_5ftype_5f_1070',['custom_type_',['../classdlaf_1_1comm_1_1_message.html#a0172d85a148bcb16f9661a76834fe384',1,'dlaf::comm::Message']]] + ['classic_5ftype_5f_1075',['classic_type_',['../classdlaf_1_1comm_1_1_message.html#a49412018d3ffdfdbd87690f59fb54a14',1,'dlaf::comm::Message']]], + ['custom_5ftype_5f_1076',['custom_type_',['../classdlaf_1_1comm_1_1_message.html#a0172d85a148bcb16f9661a76834fe384',1,'dlaf::comm::Message']]] ]; diff --git a/master/search/variables_2.js b/master/search/variables_2.js index 15de1a60ee..d0380c2558 100644 --- a/master/search/variables_2.js +++ b/master/search/variables_2.js @@ -1,6 +1,6 @@ var searchData= [ - ['data_5f_1071',['data_',['../classdlaf_1_1comm_1_1_message.html#a93caa3c42cfc16e9a27c65c37fa42c72',1,'dlaf::comm::Message::data_()'],['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a0aadac7bac5b0878d00305d21666c342',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::data_()']]], - ['dim_5f_1072',['dim_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a11827cae3d4fbece622dc9dc2d0a5014',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['dist_5fmatrix_5f_1073',['dist_matrix_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ab991456b59d36ac84dd9277edefc42ac',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] + ['data_5f_1077',['data_',['../classdlaf_1_1comm_1_1_message.html#a93caa3c42cfc16e9a27c65c37fa42c72',1,'dlaf::comm::Message::data_()'],['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a0aadac7bac5b0878d00305d21666c342',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >::data_()']]], + ['dim_5f_1078',['dim_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a11827cae3d4fbece622dc9dc2d0a5014',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['dist_5fmatrix_5f_1079',['dist_matrix_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ab991456b59d36ac84dd9277edefc42ac',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] ]; diff --git a/master/search/variables_3.js b/master/search/variables_3.js index cb425da9db..f14f1bd065 100644 --- a/master/search/variables_3.js +++ b/master/search/variables_3.js @@ -1,5 +1,5 @@ var searchData= [ - ['end_5f_1074',['end_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a6d52e171776a86d6a42b2e0cc7a26e7b',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['external_5f_1075',['external_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a785fd3e6c916a2ae30efb17f3449c230',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] + ['end_5f_1080',['end_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a6d52e171776a86d6a42b2e0cc7a26e7b',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['external_5f_1081',['external_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a785fd3e6c916a2ae30efb17f3449c230',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] ]; diff --git a/master/search/variables_4.js b/master/search/variables_4.js index aa49569f84..54cef6ca2d 100644 --- a/master/search/variables_4.js +++ b/master/search/variables_4.js @@ -1,4 +1,4 @@ var searchData= [ - ['internal_5f_1076',['internal_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#adf84f8c3806212547305082bf21ac006',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] + ['internal_5f_1082',['internal_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#adf84f8c3806212547305082bf21ac006',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] ]; diff --git a/master/search/variables_5.js b/master/search/variables_5.js index c93ed2dcbe..0de38192f8 100644 --- a/master/search/variables_5.js +++ b/master/search/variables_5.js @@ -1,5 +1,5 @@ var searchData= [ - ['start_5f_1077',['start_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ac49583acc02b6ba394d9b4e3b7450e53',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], - ['start_5foffset_5f_1078',['start_offset_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a8603ffce77527952a466a9630d98cd5c',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] + ['start_5f_1083',['start_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#ac49583acc02b6ba394d9b4e3b7450e53',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]], + ['start_5foffset_5f_1084',['start_offset_',['../structdlaf_1_1matrix_1_1_panel_3_01axis_00_01const_01_t_00_01_d_00_01_store_transposed_1_1_no_01_4.html#a8603ffce77527952a466a9630d98cd5c',1,'dlaf::matrix::Panel< axis, const T, D, StoreTransposed::No >']]] ]; diff --git a/master/solver_2triangular_8h.html b/master/solver_2triangular_8h.html index 5dd3e719ba..279e12fa82 100644 --- a/master/solver_2triangular_8h.html +++ b/master/solver_2triangular_8h.html @@ -163,6 +163,8 @@

    22 namespace dlaf {
    23 namespace solver {
    24 
    -
    42 template <Backend backend, Device device, class T>
    -
    43 void triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha,
    -
    44  Matrix<const T, device>& mat_a, Matrix<T, device>& mat_b) {
    -
    45  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    -
    46  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    -
    47  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
    -
    48  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
    -
    49 
    -
    50  if (side == blas::Side::Left) {
    -
    51  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_b, op, blas::Op::NoTrans), mat_a, mat_b, op);
    +
    43 template <Backend backend, Device device, class T>
    +
    44 void triangular(blas::Side side, blas::Uplo uplo, blas::Op op, blas::Diag diag, T alpha,
    +
    45  Matrix<const T, device>& mat_a, Matrix<T, device>& mat_b) {
    +
    46  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    +
    47  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    +
    48  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
    +
    49  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
    +
    50  DLAF_ASSERT(matrix::local_matrix(mat_a), mat_a);
    +
    51  DLAF_ASSERT(matrix::local_matrix(mat_b), mat_b);
    52 
    -
    53  if (uplo == blas::Uplo::Lower) {
    -
    54  if (op == blas::Op::NoTrans) {
    -
    55  internal::Triangular<backend, device, T>::call_LLN(diag, alpha, mat_a, mat_b);
    -
    56  }
    -
    57  else {
    -
    58  internal::Triangular<backend, device, T>::call_LLT(op, diag, alpha, mat_a, mat_b);
    +
    53  if (side == blas::Side::Left) {
    +
    54  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_b, op, blas::Op::NoTrans), mat_a, mat_b, op);
    +
    55 
    +
    56  if (uplo == blas::Uplo::Lower) {
    +
    57  if (op == blas::Op::NoTrans) {
    +
    58  internal::Triangular<backend, device, T>::call_LLN(diag, alpha, mat_a, mat_b);
    59  }
    -
    60  }
    -
    61  else {
    -
    62  if (op == blas::Op::NoTrans) {
    -
    63  internal::Triangular<backend, device, T>::call_LUN(diag, alpha, mat_a, mat_b);
    -
    64  }
    -
    65  else {
    -
    66  internal::Triangular<backend, device, T>::call_LUT(op, diag, alpha, mat_a, mat_b);
    +
    60  else {
    +
    61  internal::Triangular<backend, device, T>::call_LLT(op, diag, alpha, mat_a, mat_b);
    +
    62  }
    +
    63  }
    +
    64  else {
    +
    65  if (op == blas::Op::NoTrans) {
    +
    66  internal::Triangular<backend, device, T>::call_LUN(diag, alpha, mat_a, mat_b);
    67  }
    -
    68  }
    -
    69  }
    -
    70  else {
    -
    71  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_b, blas::Op::NoTrans, op), mat_a, mat_b, op);
    -
    72 
    -
    73  if (uplo == blas::Uplo::Lower) {
    -
    74  if (op == blas::Op::NoTrans) {
    -
    75  internal::Triangular<backend, device, T>::call_RLN(diag, alpha, mat_a, mat_b);
    -
    76  }
    -
    77  else {
    -
    78  internal::Triangular<backend, device, T>::call_RLT(op, diag, alpha, mat_a, mat_b);
    +
    68  else {
    +
    69  internal::Triangular<backend, device, T>::call_LUT(op, diag, alpha, mat_a, mat_b);
    +
    70  }
    +
    71  }
    +
    72  }
    +
    73  else {
    +
    74  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_b, blas::Op::NoTrans, op), mat_a, mat_b, op);
    +
    75 
    +
    76  if (uplo == blas::Uplo::Lower) {
    +
    77  if (op == blas::Op::NoTrans) {
    +
    78  internal::Triangular<backend, device, T>::call_RLN(diag, alpha, mat_a, mat_b);
    79  }
    -
    80  }
    -
    81  else {
    -
    82  if (op == blas::Op::NoTrans) {
    -
    83  internal::Triangular<backend, device, T>::call_RUN(diag, alpha, mat_a, mat_b);
    -
    84  }
    -
    85  else {
    -
    86  internal::Triangular<backend, device, T>::call_RUT(op, diag, alpha, mat_a, mat_b);
    +
    80  else {
    +
    81  internal::Triangular<backend, device, T>::call_RLT(op, diag, alpha, mat_a, mat_b);
    +
    82  }
    +
    83  }
    +
    84  else {
    +
    85  if (op == blas::Op::NoTrans) {
    +
    86  internal::Triangular<backend, device, T>::call_RUN(diag, alpha, mat_a, mat_b);
    87  }
    -
    88  }
    -
    89  }
    -
    90 }
    -
    91 
    -
    111 template <Backend backend, Device device, class T>
    -
    112 void triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op,
    -
    113  blas::Diag diag, T alpha, Matrix<const T, device>& mat_a, Matrix<T, device>& mat_b) {
    -
    114  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    -
    115  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    -
    116  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
    -
    117  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
    -
    118 
    -
    119  if (side == blas::Side::Left) {
    -
    120  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_b, op, blas::Op::NoTrans), mat_a, mat_b, op);
    -
    121 
    -
    122  if (uplo == blas::Uplo::Lower) {
    -
    123  if (op == blas::Op::NoTrans) {
    -
    124  internal::Triangular<backend, device, T>::call_LLN(grid, diag, alpha, mat_a, mat_b);
    -
    125  }
    -
    126  else {
    -
    127  internal::Triangular<backend, device, T>::call_LLT(grid, op, diag, alpha, mat_a, mat_b);
    -
    128  }
    -
    129  }
    -
    130  else {
    -
    131  if (op == blas::Op::NoTrans) {
    -
    132  internal::Triangular<backend, device, T>::call_LUN(grid, diag, alpha, mat_a, mat_b);
    -
    133  }
    -
    134  else {
    -
    135  internal::Triangular<backend, device, T>::call_LUT(grid, op, diag, alpha, mat_a, mat_b);
    -
    136  }
    -
    137  }
    -
    138  }
    -
    139  else {
    -
    140  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_b, blas::Op::NoTrans, op), mat_a, mat_b, op);
    -
    141 
    -
    142  if (uplo == blas::Uplo::Lower) {
    -
    143  if (op == blas::Op::NoTrans) {
    -
    144  internal::Triangular<backend, device, T>::call_RLN(grid, diag, alpha, mat_a, mat_b);
    -
    145  }
    -
    146  else {
    -
    147  internal::Triangular<backend, device, T>::call_RLT(grid, op, diag, alpha, mat_a, mat_b);
    -
    148  }
    -
    149  }
    -
    150  else {
    -
    151  if (op == blas::Op::NoTrans) {
    -
    152  internal::Triangular<backend, device, T>::call_RUN(grid, diag, alpha, mat_a, mat_b);
    -
    153  }
    -
    154  else {
    -
    155  internal::Triangular<backend, device, T>::call_RUT(grid, op, diag, alpha, mat_a, mat_b);
    -
    156  }
    -
    157  }
    -
    158  }
    -
    159 }
    -
    160 
    -
    161 }
    -
    162 }
    +
    88  else {
    +
    89  internal::Triangular<backend, device, T>::call_RUT(op, diag, alpha, mat_a, mat_b);
    +
    90  }
    +
    91  }
    +
    92  }
    +
    93 }
    +
    94 
    +
    115 template <Backend backend, Device device, class T>
    +
    116 void triangular(comm::CommunicatorGrid grid, blas::Side side, blas::Uplo uplo, blas::Op op,
    +
    117  blas::Diag diag, T alpha, Matrix<const T, device>& mat_a, Matrix<T, device>& mat_b) {
    +
    118  DLAF_ASSERT(matrix::square_size(mat_a), mat_a);
    +
    119  DLAF_ASSERT(matrix::square_blocksize(mat_a), mat_a);
    +
    120  DLAF_ASSERT(matrix::single_tile_per_block(mat_a), mat_a);
    +
    121  DLAF_ASSERT(matrix::single_tile_per_block(mat_b), mat_b);
    +
    122  DLAF_ASSERT(matrix::equal_process_grid(mat_a, grid), mat_a, grid);
    +
    123  DLAF_ASSERT(matrix::equal_process_grid(mat_b, grid), mat_b, grid);
    +
    124 
    +
    125  if (side == blas::Side::Left) {
    +
    126  DLAF_ASSERT(matrix::multipliable(mat_a, mat_b, mat_b, op, blas::Op::NoTrans), mat_a, mat_b, op);
    +
    127 
    +
    128  if (uplo == blas::Uplo::Lower) {
    +
    129  if (op == blas::Op::NoTrans) {
    +
    130  internal::Triangular<backend, device, T>::call_LLN(grid, diag, alpha, mat_a, mat_b);
    +
    131  }
    +
    132  else {
    +
    133  internal::Triangular<backend, device, T>::call_LLT(grid, op, diag, alpha, mat_a, mat_b);
    +
    134  }
    +
    135  }
    +
    136  else {
    +
    137  if (op == blas::Op::NoTrans) {
    +
    138  internal::Triangular<backend, device, T>::call_LUN(grid, diag, alpha, mat_a, mat_b);
    +
    139  }
    +
    140  else {
    +
    141  internal::Triangular<backend, device, T>::call_LUT(grid, op, diag, alpha, mat_a, mat_b);
    +
    142  }
    +
    143  }
    +
    144  }
    +
    145  else {
    +
    146  DLAF_ASSERT(matrix::multipliable(mat_b, mat_a, mat_b, blas::Op::NoTrans, op), mat_a, mat_b, op);
    +
    147 
    +
    148  if (uplo == blas::Uplo::Lower) {
    +
    149  if (op == blas::Op::NoTrans) {
    +
    150  internal::Triangular<backend, device, T>::call_RLN(grid, diag, alpha, mat_a, mat_b);
    +
    151  }
    +
    152  else {
    +
    153  internal::Triangular<backend, device, T>::call_RLT(grid, op, diag, alpha, mat_a, mat_b);
    +
    154  }
    +
    155  }
    +
    156  else {
    +
    157  if (op == blas::Op::NoTrans) {
    +
    158  internal::Triangular<backend, device, T>::call_RUN(grid, diag, alpha, mat_a, mat_b);
    +
    159  }
    +
    160  else {
    +
    161  internal::Triangular<backend, device, T>::call_RUT(grid, op, diag, alpha, mat_a, mat_b);
    +
    162  }
    +
    163  }
    +
    164  }
    +
    165 }
    +
    166 
    +
    167 }
    +
    168 }
    dlaf::comm::CommunicatorGrid
    Definition: communicator_grid.h:42
    communicator_grid.h
    diff --git a/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_563250733b77a55bd996211659d75e42.html b/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_563250733b77a55bd996211659d75e42.html index 4ee7608d74..8ead579438 100644 --- a/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_563250733b77a55bd996211659d75e42.html +++ b/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_563250733b77a55bd996211659d75e42.html @@ -72,8 +72,8 @@

    This is the complete list of members for dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >, including all inherited members.

    - - + + diff --git a/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_e566a066992e6980ee13cab3c8831bc5.html b/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_e566a066992e6980ee13cab3c8831bc5.html index 617f67a71d..a5e8920e2e 100644 --- a/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_e566a066992e6980ee13cab3c8831bc5.html +++ b/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_g_p_e566a066992e6980ee13cab3c8831bc5.html @@ -80,13 +80,13 @@ - - - + + - - + +
    call(Matrix< T, Device::GPU > &mat_a, matrix::RetiledMatrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >inline
    call(TriggerSender &&trigger, comm::IndexT_MPI rank_v0, CommSender &&mpi_col_chain_panel, Matrix< T, D > &mat_a, matrix::RetiledMatrix< T, Device::CPU > &mat_taus, SizeType j_sub, const matrix::SubPanelView &panel_view) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >inline
    call(Matrix< T, Device::GPU > &mat_a, Matrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >inline
    call(TriggerSender &&trigger, comm::IndexT_MPI rank_v0, CommSender &&mpi_col_chain_panel, Matrix< T, D > &mat_a, Matrix< T, Device::CPU > &mat_taus, SizeType j_sub, const matrix::SubPanelView &panel_view) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >inline
    ComputePanelHelper(const std::size_t n_workspaces, matrix::Distribution dist_a) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >inline
    copyFromCPU(const matrix::SubPanelView panel_view, matrix::Panel< Coord::Col, T, Device::CPU > &v, matrix::Matrix< T, Device::GPU > &mat_a) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >inlineprotected
    copyToCPU(const matrix::SubPanelView panel_view, matrix::Matrix< T, Device::GPU > &mat_a, matrix::Panel< Coord::Col, T, Device::CPU > &v) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::GPU, Device::GPU, T >inlineprotected
     ComputePanelHelper (const std::size_t n_workspaces, matrix::Distribution dist_a)
     
    -void call (Matrix< T, Device::GPU > &mat_a, matrix::RetiledMatrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view)
     
    +
    +void call (Matrix< T, Device::GPU > &mat_a, Matrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view)
     
    template<Device D, class CommSender , class TriggerSender >
    void call (TriggerSender &&trigger, comm::IndexT_MPI rank_v0, CommSender &&mpi_col_chain_panel, Matrix< T, D > &mat_a, matrix::RetiledMatrix< T, Device::CPU > &mat_taus, SizeType j_sub, const matrix::SubPanelView &panel_view)
     
    void call (TriggerSender &&trigger, comm::IndexT_MPI rank_v0, CommSender &&mpi_col_chain_panel, Matrix< T, D > &mat_a, Matrix< T, Device::CPU > &mat_taus, SizeType j_sub, const matrix::SubPanelView &panel_view)
     
    diff --git a/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_9c17309f77611319159a292eeae019d5.html b/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_9c17309f77611319159a292eeae019d5.html index 7066d7db16..8f154cc84a 100644 --- a/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_9c17309f77611319159a292eeae019d5.html +++ b/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_9c17309f77611319159a292eeae019d5.html @@ -72,8 +72,8 @@

    This is the complete list of members for dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >, including all inherited members.

    Protected Member Functions

    - - + +
    call(Matrix< T, Device::CPU > &mat_a, matrix::RetiledMatrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >inline
    call(TriggerSender &&trigger, comm::IndexT_MPI rank_v0, CommSender &&mpi_col_chain_panel, Matrix< T, D > &mat_a, matrix::RetiledMatrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >inline
    call(Matrix< T, Device::CPU > &mat_a, Matrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >inline
    call(TriggerSender &&trigger, comm::IndexT_MPI rank_v0, CommSender &&mpi_col_chain_panel, Matrix< T, D > &mat_a, Matrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >inline
    ComputePanelHelper(const std::size_t, matrix::Distribution) (defined in dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >)dlaf::eigensolver::internal::red2band::ComputePanelHelper< Backend::MC, Device::CPU, T >inline
    diff --git a/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_f6049763c467d19a4a2bae31f1d19649.html b/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_f6049763c467d19a4a2bae31f1d19649.html index 865359e11d..6a3479e179 100644 --- a/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_f6049763c467d19a4a2bae31f1d19649.html +++ b/master/structdlaf_1_1eigensolver_1_1internal_1_1red2band_1_1_compute_panel_helper_3_01_backend_1_1_m_c_f6049763c467d19a4a2bae31f1d19649.html @@ -78,13 +78,13 @@  ComputePanelHelper (const std::size_t, matrix::Distribution)   - -void call (Matrix< T, Device::CPU > &mat_a, matrix::RetiledMatrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) -  - + +void call (Matrix< T, Device::CPU > &mat_a, Matrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) +  + template<Device D, class CommSender , class TriggerSender > -void call (TriggerSender &&trigger, comm::IndexT_MPI rank_v0, CommSender &&mpi_col_chain_panel, Matrix< T, D > &mat_a, matrix::RetiledMatrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) -  +void call (TriggerSender &&trigger, comm::IndexT_MPI rank_v0, CommSender &&mpi_col_chain_panel, Matrix< T, D > &mat_a, Matrix< T, Device::CPU > &mat_taus, const SizeType j_sub, const matrix::SubPanelView &panel_view) + 
    The documentation for this struct was generated from the following file:
    • /home/runner/work/DLA-Future/DLA-Future/include/dlaf/eigensolver/reduction_to_band/impl.h
      • diff --git a/master/tridiag__solver_8h.html b/master/tridiag__solver_8h.html index adb2a728e9..64fe99c2ed 100644 --- a/master/tridiag__solver_8h.html +++ b/master/tridiag__solver_8h.html @@ -142,9 +142,15 @@

      21 namespace dlaf {
      22 namespace eigensolver {
      23 
      -
      38 template <Backend backend, Device device, class T>
      -
      39 void tridiagSolver(Matrix<BaseType<T>, Device::CPU>& tridiag, Matrix<BaseType<T>, device>& evals,
      -
      40  Matrix<T, device>& evecs) {
      -
      41  DLAF_ASSERT(matrix::local_matrix(tridiag), tridiag);
      -
      42  DLAF_ASSERT(tridiag.distribution().size().cols() == 2, tridiag);
      -
      43  DLAF_ASSERT(tridiag.distribution().blockSize().cols() == 2, tridiag);
      -
      44 
      -
      45  DLAF_ASSERT(matrix::local_matrix(evals), evals);
      -
      46  DLAF_ASSERT(evals.distribution().size().cols() == 1, evals);
      -
      47 
      -
      48  DLAF_ASSERT(matrix::local_matrix(evecs), evecs);
      -
      49  DLAF_ASSERT(matrix::square_size(evecs), evecs);
      -
      50  DLAF_ASSERT(matrix::square_blocksize(evecs), evecs);
      +
      41 template <Backend backend, Device device, class T>
      +
      42 void tridiagSolver(Matrix<BaseType<T>, Device::CPU>& tridiag, Matrix<BaseType<T>, device>& evals,
      +
      43  Matrix<T, device>& evecs) {
      +
      44  DLAF_ASSERT(matrix::local_matrix(tridiag), tridiag);
      +
      45  DLAF_ASSERT(tridiag.distribution().size().cols() == 2, tridiag);
      +
      46  DLAF_ASSERT(tridiag.distribution().blockSize().cols() == 2, tridiag);
      +
      47  DLAF_ASSERT(matrix::single_tile_per_block(tridiag), tridiag);
      +
      48 
      +
      49  DLAF_ASSERT(matrix::local_matrix(evals), evals);
      +
      50  DLAF_ASSERT(evals.distribution().size().cols() == 1, evals);
      51 
      -
      52  DLAF_ASSERT(tridiag.distribution().blockSize().rows() == evecs.distribution().blockSize().rows(),
      -
      53  evecs.distribution().blockSize().rows(), tridiag.distribution().blockSize().rows());
      -
      54  DLAF_ASSERT(tridiag.distribution().blockSize().rows() == evals.distribution().blockSize().rows(),
      -
      55  tridiag.distribution().blockSize().rows(), evals.distribution().blockSize().rows());
      -
      56  DLAF_ASSERT(tridiag.distribution().size().rows() == evecs.distribution().size().rows(),
      -
      57  evecs.distribution().size().rows(), tridiag.distribution().size().rows());
      -
      58  DLAF_ASSERT(tridiag.distribution().size().rows() == evals.distribution().size().rows(),
      -
      59  tridiag.distribution().size().rows(), evals.distribution().size().rows());
      -
      60 
      -
      61  internal::TridiagSolver<backend, device, BaseType<T>>::call(tridiag, evals, evecs);
      -
      62 }
      -
      63 
      -
      80 template <Backend B, Device D, class T>
      -
      81 void tridiagSolver(comm::CommunicatorGrid grid, Matrix<BaseType<T>, Device::CPU>& tridiag,
      -
      82  Matrix<BaseType<T>, D>& evals, Matrix<T, D>& evecs) {
      -
      83  DLAF_ASSERT(matrix::local_matrix(tridiag), tridiag);
      -
      84  DLAF_ASSERT(tridiag.distribution().size().cols() == 2, tridiag);
      -
      85  DLAF_ASSERT(tridiag.distribution().blockSize().cols() == 2, tridiag);
      -
      86 
      -
      87  DLAF_ASSERT(matrix::local_matrix(evals), evals);
      -
      88  DLAF_ASSERT(evals.distribution().size().cols() == 1, evals);
      -
      89 
      -
      90  DLAF_ASSERT(matrix::square_size(evecs), evecs);
      -
      91  DLAF_ASSERT(matrix::square_blocksize(evecs), evecs);
      -
      92  DLAF_ASSERT(matrix::equal_process_grid(evecs, grid), evecs, grid);
      -
      93 
      -
      94  DLAF_ASSERT(tridiag.distribution().blockSize().rows() == evecs.distribution().blockSize().rows(),
      -
      95  evecs, tridiag);
      -
      96  DLAF_ASSERT(tridiag.distribution().blockSize().rows() == evals.distribution().blockSize().rows(),
      -
      97  tridiag, evals);
      -
      98  DLAF_ASSERT(tridiag.distribution().size().rows() == evecs.distribution().size().rows(), evecs,
      -
      99  tridiag);
      -
      100  DLAF_ASSERT(tridiag.distribution().size().rows() == evals.distribution().size().rows(), tridiag,
      -
      101  evals);
      -
      102 
      -
      103  internal::TridiagSolver<B, D, BaseType<T>>::call(grid, tridiag, evals, evecs);
      -
      104 }
      -
      105 
      -
      106 }
      -
      107 }
      +
      52  DLAF_ASSERT(matrix::local_matrix(evecs), evecs);
      +
      53  DLAF_ASSERT(matrix::square_size(evecs), evecs);
      +
      54  DLAF_ASSERT(matrix::square_blocksize(evecs), evecs);
      +
      55 
      +
      56  DLAF_ASSERT(matrix::single_tile_per_block(evecs), evecs);
      +
      57  DLAF_ASSERT(matrix::single_tile_per_block(evals), evals);
      +
      58 
      +
      59  DLAF_ASSERT(tridiag.distribution().blockSize().rows() == evecs.distribution().blockSize().rows(),
      +
      60  evecs.distribution().blockSize().rows(), tridiag.distribution().blockSize().rows());
      +
      61  DLAF_ASSERT(tridiag.distribution().blockSize().rows() == evals.distribution().blockSize().rows(),
      +
      62  tridiag.distribution().blockSize().rows(), evals.distribution().blockSize().rows());
      +
      63  DLAF_ASSERT(tridiag.distribution().size().rows() == evecs.distribution().size().rows(),
      +
      64  evecs.distribution().size().rows(), tridiag.distribution().size().rows());
      +
      65  DLAF_ASSERT(tridiag.distribution().size().rows() == evals.distribution().size().rows(),
      +
      66  tridiag.distribution().size().rows(), evals.distribution().size().rows());
      +
      67 
      +
      68  internal::TridiagSolver<backend, device, BaseType<T>>::call(tridiag, evals, evecs);
      +
      69 }
      +
      70 
      +
      93 template <Backend B, Device D, class T>
      +
      94 void tridiagSolver(comm::CommunicatorGrid grid, Matrix<BaseType<T>, Device::CPU>& tridiag,
      +
      95  Matrix<BaseType<T>, D>& evals, Matrix<T, D>& evecs) {
      +
      96  DLAF_ASSERT(matrix::local_matrix(tridiag), tridiag);
      +
      97  DLAF_ASSERT(tridiag.distribution().size().cols() == 2, tridiag);
      +
      98  DLAF_ASSERT(tridiag.distribution().blockSize().cols() == 2, tridiag);
      +
      99  DLAF_ASSERT(matrix::single_tile_per_block(tridiag), tridiag);
      +
      100 
      +
      101  DLAF_ASSERT(matrix::local_matrix(evals), evals);
      +
      102  DLAF_ASSERT(evals.distribution().size().cols() == 1, evals);
      +
      103 
      +
      104  DLAF_ASSERT(matrix::square_size(evecs), evecs);
      +
      105  DLAF_ASSERT(matrix::square_blocksize(evecs), evecs);
      +
      106  DLAF_ASSERT(matrix::equal_process_grid(evecs, grid), evecs, grid);
      +
      107 
      +
      108  DLAF_ASSERT(matrix::single_tile_per_block(evecs), evecs);
      +
      109  DLAF_ASSERT(matrix::single_tile_per_block(evals), evals);
      +
      110 
      +
      111  DLAF_ASSERT(tridiag.distribution().blockSize().rows() == evecs.distribution().blockSize().rows(),
      +
      112  evecs, tridiag);
      +
      113  DLAF_ASSERT(tridiag.distribution().blockSize().rows() == evals.distribution().blockSize().rows(),
      +
      114  tridiag, evals);
      +
      115  DLAF_ASSERT(tridiag.distribution().size().rows() == evecs.distribution().size().rows(), evecs,
      +
      116  tridiag);
      +
      117  DLAF_ASSERT(tridiag.distribution().size().rows() == evals.distribution().size().rows(), tridiag,
      +
      118  evals);
      +
      119 
      +
      120  internal::TridiagSolver<B, D, BaseType<T>>::call(grid, tridiag, evals, evecs);
      +
      121 }
      +
      122 
      +
      123 }
      +
      124 }
      dlaf::comm::CommunicatorGrid
      Definition: communicator_grid.h:42
      communicator_grid.h
      dlaf::eigensolver::eigensolver
      void eigensolver(blas::Uplo uplo, Matrix< T, D > &mat, Matrix< BaseType< T >, D > &eigenvalues, Matrix< T, D > &eigenvectors)
      Definition: eigensolver.h:39
      -
      dlaf::eigensolver::tridiagSolver
      void tridiagSolver(Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, device > &evals, Matrix< T, device > &evecs)
      Definition: tridiag_solver.h:39
      +
      dlaf::eigensolver::tridiagSolver
      void tridiagSolver(Matrix< BaseType< T >, Device::CPU > &tridiag, Matrix< BaseType< T >, device > &evals, Matrix< T, device > &evecs)
      Definition: tridiag_solver.h:42
      util_matrix.h
      diff --git a/master/util__matrix_8h.html b/master/util__matrix_8h.html index 6a7371ef33..383ed5b331 100644 --- a/master/util__matrix_8h.html +++ b/master/util__matrix_8h.html @@ -112,6 +112,11 @@ bool dlaf::matrix::square_blocksize (const MatrixLike &m) noexcept  Returns true if the matrix block size is square.
        + +template<class MatrixLike > +bool dlaf::matrix::single_tile_per_block (const MatrixLike &m) noexcept + Returns true if the matrix has a single tile per block.
      +  template<class MatrixLikeA , class MatrixLikeB > bool dlaf::matrix::equal_size (const MatrixLikeA &lhs, const MatrixLikeB &rhs) noexcept diff --git a/master/util__matrix_8h_source.html b/master/util__matrix_8h_source.html index a3a3cd1b24..f7a0dec5bc 100644 --- a/master/util__matrix_8h_source.html +++ b/master/util__matrix_8h_source.html @@ -117,272 +117,277 @@
      49  return m.blockSize().rows() == m.blockSize().cols();
      50 }
      51 
      -
      53 template <class MatrixLikeA, class MatrixLikeB>
      -
      54 bool equal_size(const MatrixLikeA& lhs, const MatrixLikeB& rhs) noexcept {
      -
      55  return lhs.size() == rhs.size();
      +
      53 template <class MatrixLike>
      +
      54 bool single_tile_per_block(const MatrixLike& m) noexcept {
      +
      55  return m.blockSize() == m.baseTileSize();
      56 }
      57 
      -
      59 template <class T, Device D1, Device D2>
      -
      60 bool equal_blocksize(const Matrix<const T, D1>& lhs, Matrix<const T, D2>& rhs) noexcept {
      -
      61  return lhs.blockSize() == rhs.blockSize();
      +
      59 template <class MatrixLikeA, class MatrixLikeB>
      +
      60 bool equal_size(const MatrixLikeA& lhs, const MatrixLikeB& rhs) noexcept {
      +
      61  return lhs.size() == rhs.size();
      62 }
      63 
      -
      65 template <class T, Device D>
      -
      66 bool local_matrix(const Matrix<const T, D>& m) noexcept {
      -
      67  return m.commGridSize() == comm::Size2D(1, 1);
      +
      65 template <class T, Device D1, Device D2>
      +
      66 bool equal_blocksize(const Matrix<const T, D1>& lhs, Matrix<const T, D2>& rhs) noexcept {
      +
      67  return lhs.blockSize() == rhs.blockSize();
      68 }
      69 
      71 template <class T, Device D>
      -
      72 bool equal_process_grid(const Matrix<const T, D>& m, const comm::CommunicatorGrid& g) noexcept {
      -
      73  return m.commGridSize() == g.size() && m.rankIndex() == g.rank();
      +
      72 bool local_matrix(const Matrix<const T, D>& m) noexcept {
      +
      73  return m.commGridSize() == comm::Size2D(1, 1);
      74 }
      75 
      -
      77 template <class T, Device D1, Device D2>
      -
      78 bool equal_distributions(const Matrix<const T, D1>& lhs, const Matrix<const T, D2>& rhs) noexcept {
      -
      79  return lhs.distribution() == rhs.distribution();
      +
      77 template <class T, Device D>
      +
      78 bool equal_process_grid(const Matrix<const T, D>& m, const comm::CommunicatorGrid& g) noexcept {
      +
      79  return m.commGridSize() == g.size() && m.rankIndex() == g.rank();
      80 }
      81 
      -
      83 template <class IndexT, class Tag>
      -
      84 bool multipliable_sizes(common::Size2D<IndexT, Tag> a, common::Size2D<IndexT, Tag> b,
      -
      85  common::Size2D<IndexT, Tag> c, const blas::Op opA, const blas::Op opB) noexcept {
      -
      86  if (opA != blas::Op::NoTrans)
      -
      87  a.transpose();
      -
      88  if (opB != blas::Op::NoTrans)
      -
      89  b.transpose();
      -
      90 
      -
      91  return a.rows() == c.rows() && a.cols() == b.rows() && b.cols() == c.cols();
      -
      92 }
      -
      93 
      -
      95 template <class T, Device D>
      -
      96 bool multipliable(const Matrix<const T, D>& a, const Matrix<const T, D>& b, const Matrix<const T, D>& c,
      -
      97  const blas::Op opA, const blas::Op opB) noexcept {
      -
      98  return multipliable_sizes(a.size(), b.size(), c.size(), opA, opB) &&
      -
      99  multipliable_sizes(a.blockSize(), b.blockSize(), c.blockSize(), opA, opB);
      -
      100 }
      -
      101 
      -
      102 namespace util {
      -
      103 namespace internal {
      -
      104 
      -
      106 template <class T>
      -
      107 class getter_random {
      -
      108  static_assert(std::is_same_v<T, float> || std::is_same_v<T, double>,
      -
      109  "T is not compatible with random generator used.");
      +
      83 template <class T, Device D1, Device D2>
      +
      84 bool equal_distributions(const Matrix<const T, D1>& lhs, const Matrix<const T, D2>& rhs) noexcept {
      +
      85  return lhs.distribution() == rhs.distribution();
      +
      86 }
      +
      87 
      +
      89 template <class IndexT, class Tag>
      +
      90 bool multipliable_sizes(common::Size2D<IndexT, Tag> a, common::Size2D<IndexT, Tag> b,
      +
      91  common::Size2D<IndexT, Tag> c, const blas::Op opA, const blas::Op opB) noexcept {
      +
      92  if (opA != blas::Op::NoTrans)
      +
      93  a.transpose();
      +
      94  if (opB != blas::Op::NoTrans)
      +
      95  b.transpose();
      +
      96 
      +
      97  return a.rows() == c.rows() && a.cols() == b.rows() && b.cols() == c.cols();
      +
      98 }
      +
      99 
      +
      101 template <class T, Device D>
      +
      102 bool multipliable(const Matrix<const T, D>& a, const Matrix<const T, D>& b, const Matrix<const T, D>& c,
      +
      103  const blas::Op opA, const blas::Op opB) noexcept {
      +
      104  return multipliable_sizes(a.size(), b.size(), c.size(), opA, opB) &&
      +
      105  multipliable_sizes(a.blockSize(), b.blockSize(), c.blockSize(), opA, opB);
      +
      106 }
      +
      107 
      +
      108 namespace util {
      +
      109 namespace internal {
      110 
      -
      111 public:
      -
      112  getter_random() : random_engine_(std::minstd_rand::default_seed) {}
      -
      113  getter_random(SizeType seed) : random_engine_(static_cast<std::size_t>(seed)) {
      -
      114  DLAF_ASSERT(seed >= 0, "");
      -
      115  }
      +
      112 template <class T>
      +
      113 class getter_random {
      +
      114  static_assert(std::is_same_v<T, float> || std::is_same_v<T, double>,
      +
      115  "T is not compatible with random generator used.");
      116 
      -
      117  T operator()() {
      -
      118  return random_sampler_(random_engine_);
      -
      119  }
      -
      120 
      -
      121 private:
      -
      122  std::mt19937_64 random_engine_;
      -
      123  std::uniform_real_distribution<T> random_sampler_{-1, 1};
      -
      124 };
      -
      125 
      -
      127 template <class T>
      -
      128 class getter_random<std::complex<T>> : private getter_random<T> {
      -
      129 public:
      -
      130  using getter_random<T>::getter_random;
      +
      117 public:
      +
      118  getter_random() : random_engine_(std::minstd_rand::default_seed) {}
      +
      119  getter_random(SizeType seed) : random_engine_(static_cast<std::size_t>(seed)) {
      +
      120  DLAF_ASSERT(seed >= 0, "");
      +
      121  }
      +
      122 
      +
      123  T operator()() {
      +
      124  return random_sampler_(random_engine_);
      +
      125  }
      +
      126 
      +
      127 private:
      +
      128  std::mt19937_64 random_engine_;
      +
      129  std::uniform_real_distribution<T> random_sampler_{-1, 1};
      +
      130 };
      131 
      -
      132  std::complex<T> operator()() {
      -
      133  return std::polar<T>(std::abs(getter_random<T>::operator()()),
      -
      134  static_cast<T>(M_PI) * getter_random<T>::operator()());
      -
      135  }
      -
      136 };
      +
      133 template <class T>
      +
      134 class getter_random<std::complex<T>> : private getter_random<T> {
      +
      135 public:
      +
      136  using getter_random<T>::getter_random;
      137 
      -
      138 }
      -
      139 
      -
      142 template <Backend backend, class T, Device D>
      -
      143 void set0(pika::execution::thread_priority priority, LocalTileIndex begin, LocalTileSize sz,
      -
      144  Matrix<T, D>& matrix) {
      -
      145  using dlaf::internal::Policy;
      -
      146  using pika::execution::experimental::start_detached;
      -
      147 
      -
      148  for (const auto& idx : iterate_range2d(begin, sz))
      -
      149  start_detached(matrix.readwrite(idx) | tile::set0(Policy<backend>(priority)));
      -
      150 }
      -
      151 
      -
      156 template <Backend backend, class T, Device D>
      -
      157 void set0(pika::execution::thread_priority priority, Matrix<T, D>& matrix) {
      -
      158  set0<backend>(priority, LocalTileIndex(0, 0), matrix.distribution().localNrTiles(), matrix);
      -
      159 }
      -
      160 
      -
      162 template <Backend backend, class T, Coord axis, Device D, StoreTransposed storage>
      -
      163 void set0(pika::execution::thread_priority priority, Panel<axis, T, D, storage>& panel) {
      -
      164  using dlaf::internal::Policy;
      -
      165  using pika::execution::experimental::start_detached;
      +
      138  std::complex<T> operator()() {
      +
      139  return std::polar<T>(std::abs(getter_random<T>::operator()()),
      +
      140  static_cast<T>(M_PI) * getter_random<T>::operator()());
      +
      141  }
      +
      142 };
      +
      143 
      +
      144 }
      +
      145 
      +
      148 template <Backend backend, class T, Device D>
      +
      149 void set0(pika::execution::thread_priority priority, LocalTileIndex begin, LocalTileSize sz,
      +
      150  Matrix<T, D>& matrix) {
      +
      151  using dlaf::internal::Policy;
      +
      152  using pika::execution::experimental::start_detached;
      +
      153 
      +
      154  for (const auto& idx : iterate_range2d(begin, sz))
      +
      155  start_detached(matrix.readwrite(idx) | tile::set0(Policy<backend>(priority)));
      +
      156 }
      +
      157 
      +
      162 template <Backend backend, class T, Device D>
      +
      163 void set0(pika::execution::thread_priority priority, Matrix<T, D>& matrix) {
      +
      164  set0<backend>(priority, LocalTileIndex(0, 0), matrix.distribution().localNrTiles(), matrix);
      +
      165 }
      166 
      -
      167  for (const auto& tile_idx : panel.iteratorLocal())
      -
      168  start_detached(panel.readwrite(tile_idx) | tile::set0(Policy<backend>(priority)));
      -
      169 }
      -
      170 
      -
      177 template <class T, class ElementGetter>
      -
      178 void set(Matrix<T, Device::CPU>& matrix, ElementGetter el_f) {
      -
      179  const Distribution& dist = matrix.distribution();
      -
      180  for (auto tile_wrt_local : iterate_range2d(dist.localNrTiles())) {
      -
      181  GlobalTileIndex tile_wrt_global = dist.globalTileIndex(tile_wrt_local);
      -
      182  auto tl_index = dist.globalElementIndex(tile_wrt_global, {0, 0});
      -
      183 
      -
      184  using TileType = typename std::decay_t<decltype(matrix)>::TileType;
      -
      185  auto set_f = [tl_index, el_f = el_f](const TileType& tile) {
      -
      186  for (auto el_idx_l : iterate_range2d(tile.size())) {
      -
      187  GlobalElementIndex el_idx_g(el_idx_l.row() + tl_index.row(), el_idx_l.col() + tl_index.col());
      -
      188  tile(el_idx_l) = el_f(el_idx_g);
      -
      189  }
      -
      190  };
      -
      191 
      -
      192  dlaf::internal::transformDetach(dlaf::internal::Policy<Backend::MC>(), std::move(set_f),
      -
      193  matrix.readwrite(tile_wrt_local));
      -
      194  }
      -
      195 }
      -
      196 
      -
      206 template <class T, class ElementGetter>
      -
      207 void set(Matrix<T, Device::CPU>& matrix, ElementGetter el_f, const blas::Op op) {
      -
      208  auto el_op_f = [op, el_f](const GlobalElementIndex& index) -> T {
      -
      209  using blas::Op;
      -
      210  switch (op) {
      -
      211  case Op::NoTrans:
      -
      212  return el_f(index);
      -
      213  case Op::Trans:
      -
      214  return el_f(transposed(index));
      -
      215  case Op::ConjTrans:
      -
      216  return dlaf::conj(el_f(transposed(index)));
      -
      217  default:
      -
      218  DLAF_UNIMPLEMENTED(op);
      -
      219  return T{};
      -
      220  }
      -
      221  };
      -
      222 
      -
      223  set(matrix, el_op_f);
      -
      224 }
      -
      225 
      -
      236 template <class T>
      -
      237 void set_random(Matrix<T, Device::CPU>& matrix) {
      -
      238  const Distribution& dist = matrix.distribution();
      -
      239  for (auto tile_wrt_local : iterate_range2d(dist.localNrTiles())) {
      -
      240  GlobalTileIndex tile_wrt_global = dist.globalTileIndex(tile_wrt_local);
      -
      241  auto tl_index = dist.globalElementIndex(tile_wrt_global, {0, 0});
      -
      242  auto seed = tl_index.col() + tl_index.row() * matrix.size().cols();
      -
      243 
      -
      244  using TileType = typename std::decay_t<decltype(matrix)>::TileType;
      -
      245  auto rnd_f = [seed](TileType&& tile) {
      -
      246  internal::getter_random<T> random_value(seed);
      -
      247  for (auto el_idx : iterate_range2d(tile.size())) {
      -
      248  tile(el_idx) = random_value();
      -
      249  }
      -
      250  };
      -
      251 
      -
      252  dlaf::internal::transformDetach(dlaf::internal::Policy<Backend::MC>(), std::move(rnd_f),
      -
      253  matrix.readwrite(tile_wrt_local));
      -
      254  }
      -
      255 }
      -
      256 
      -
      257 namespace internal {
      -
      258 
      -
      259 template <class T>
      -
      260 void set_diagonal_tile(const Tile<T, Device::CPU>& tile, internal::getter_random<T>& random_value,
      -
      261  SizeType offset_value) {
      -
      262  // DIAGONAL
      -
      263  // for diagonal tiles get just lower matrix values and set value for both
      -
      264  // straight and transposed indices
      -
      265  for (SizeType j = 0; j < tile.size().cols(); ++j) {
      -
      266  for (SizeType i = 0; i < j; ++i) {
      -
      267  auto value = random_value();
      -
      268 
      -
      269  tile(TileElementIndex{i, j}) = value;
      -
      270  tile(TileElementIndex{j, i}) = dlaf::conj(value);
      -
      271  }
      -
      272  tile(TileElementIndex{j, j}) = std::real(random_value()) + offset_value;
      -
      273  }
      -
      274 }
      -
      275 
      -
      276 template <class T>
      -
      277 void set_lower_and_upper_tile(const Tile<T, Device::CPU>& tile, internal::getter_random<T>& random_value,
      -
      278  TileElementSize full_tile_size, GlobalTileIndex tile_wrt_global) {
      -
      279  // LOWER or UPPER (except DIAGONAL)
      -
      280  // random values are requested in the same order for both original and transposed
      -
      281  for (SizeType j = 0; j < full_tile_size.cols(); ++j) {
      -
      282  for (SizeType i = 0; i < full_tile_size.rows(); ++i) {
      -
      283  auto value = random_value();
      -
      284 
      -
      285  // but they are set row-wise in the original tile and col-wise in the
      -
      286  // transposed one
      -
      287  if (tile_wrt_global.row() > tile_wrt_global.col()) { // LOWER
      -
      288  TileElementIndex index{i, j};
      -
      289  if (index.isIn(tile.size()))
      -
      290  tile(index) = value;
      -
      291  }
      -
      292  else { // UPPER
      -
      293  TileElementIndex index{j, i};
      -
      294  if (index.isIn(tile.size()))
      -
      295  tile(index) = dlaf::conj(value);
      -
      296  }
      -
      297  }
      -
      298  }
      -
      299 }
      -
      300 
      -
      317 template <class T>
      -
      318 void set_random_hermitian_with_offset(Matrix<T, Device::CPU>& matrix, const SizeType offset_value) {
      -
      319  // note:
      -
      320  // By assuming square blocksizes, it is easier to locate elements. In fact:
      -
      321  // - Elements on the diagonal are stored in the diagonal of the diagonal tiles
      -
      322  // - Tiles under the diagonal store elements of the lower triangular matrix
      -
      323  // - Tiles over the diagonal store elements of the upper triangular matrix
      -
      324 
      -
      325  const Distribution& dist = matrix.distribution();
      -
      326 
      -
      327  DLAF_ASSERT(square_size(matrix), matrix);
      -
      328  DLAF_ASSERT(square_blocksize(matrix), matrix);
      -
      329 
      -
      330  auto full_tile_size = matrix.blockSize();
      -
      331 
      -
      332  for (auto tile_wrt_local : iterate_range2d(dist.localNrTiles())) {
      -
      333  GlobalTileIndex tile_wrt_global = dist.globalTileIndex(tile_wrt_local);
      -
      334 
      -
      335  auto tl_index = dist.globalElementIndex(tile_wrt_global, {0, 0});
      -
      336 
      -
      337  // compute the same seed for original and "transposed" tiles, so transposed ones will know the
      -
      338  // values of the original one without the need of accessing real values (nor communication in case
      -
      339  // of distributed matrices)
      -
      340  SizeType seed;
      -
      341  if (tile_wrt_global.row() >= tile_wrt_global.col()) // LOWER or DIAGONAL
      -
      342  seed = tl_index.col() + tl_index.row() * matrix.size().cols();
      -
      343  else
      -
      344  seed = tl_index.row() + tl_index.col() * matrix.size().rows();
      -
      345 
      -
      346  using TileType = typename std::decay_t<decltype(matrix)>::TileType;
      -
      347  auto set_hp_f = [=](const TileType& tile) {
      -
      348  internal::getter_random<T> random_value(seed);
      -
      349  if (tile_wrt_global.row() == tile_wrt_global.col())
      -
      350  internal::set_diagonal_tile(tile, random_value, offset_value);
      -
      351  else
      -
      352  internal::set_lower_and_upper_tile(tile, random_value, full_tile_size, tile_wrt_global);
      -
      353  };
      -
      354 
      -
      355  dlaf::internal::transformDetach(dlaf::internal::Policy<Backend::MC>(), std::move(set_hp_f),
      -
      356  matrix.readwrite(tile_wrt_local));
      -
      357  }
      -
      358 }
      -
      359 
      -
      360 }
      -
      361 
      -
      375 template <class T>
      -
      376 void set_random_hermitian(Matrix<T, Device::CPU>& matrix) {
      -
      377  internal::set_random_hermitian_with_offset(matrix, 0);
      -
      378 }
      -
      379 
      -
      396 template <class T>
      -
      397 void set_random_hermitian_positive_definite(Matrix<T, Device::CPU>& matrix) {
      -
      398  internal::set_random_hermitian_with_offset(matrix, 2 * matrix.size().rows());
      -
      399 }
      -
      400 
      -
      401 }
      -
      402 }
      -
      403 }
      +
      168 template <Backend backend, class T, Coord axis, Device D, StoreTransposed storage>
      +
      169 void set0(pika::execution::thread_priority priority, Panel<axis, T, D, storage>& panel) {
      +
      170  using dlaf::internal::Policy;
      +
      171  using pika::execution::experimental::start_detached;
      +
      172 
      +
      173  for (const auto& tile_idx : panel.iteratorLocal())
      +
      174  start_detached(panel.readwrite(tile_idx) | tile::set0(Policy<backend>(priority)));
      +
      175 }
      +
      176 
      +
      183 template <class T, class ElementGetter>
      +
      184 void set(Matrix<T, Device::CPU>& matrix, ElementGetter el_f) {
      +
      185  const Distribution& dist = matrix.distribution();
      +
      186  for (auto tile_wrt_local : iterate_range2d(dist.localNrTiles())) {
      +
      187  GlobalTileIndex tile_wrt_global = dist.globalTileIndex(tile_wrt_local);
      +
      188  auto tl_index = dist.globalElementIndex(tile_wrt_global, {0, 0});
      +
      189 
      +
      190  using TileType = typename std::decay_t<decltype(matrix)>::TileType;
      +
      191  auto set_f = [tl_index, el_f = el_f](const TileType& tile) {
      +
      192  for (auto el_idx_l : iterate_range2d(tile.size())) {
      +
      193  GlobalElementIndex el_idx_g(el_idx_l.row() + tl_index.row(), el_idx_l.col() + tl_index.col());
      +
      194  tile(el_idx_l) = el_f(el_idx_g);
      +
      195  }
      +
      196  };
      +
      197 
      +
      198  dlaf::internal::transformDetach(dlaf::internal::Policy<Backend::MC>(), std::move(set_f),
      +
      199  matrix.readwrite(tile_wrt_local));
      +
      200  }
      +
      201 }
      +
      202 
      +
      212 template <class T, class ElementGetter>
      +
      213 void set(Matrix<T, Device::CPU>& matrix, ElementGetter el_f, const blas::Op op) {
      +
      214  auto el_op_f = [op, el_f](const GlobalElementIndex& index) -> T {
      +
      215  using blas::Op;
      +
      216  switch (op) {
      +
      217  case Op::NoTrans:
      +
      218  return el_f(index);
      +
      219  case Op::Trans:
      +
      220  return el_f(transposed(index));
      +
      221  case Op::ConjTrans:
      +
      222  return dlaf::conj(el_f(transposed(index)));
      +
      223  default:
      +
      224  DLAF_UNIMPLEMENTED(op);
      +
      225  return T{};
      +
      226  }
      +
      227  };
      +
      228 
      +
      229  set(matrix, el_op_f);
      +
      230 }
      +
      231 
      +
      242 template <class T>
      +
      243 void set_random(Matrix<T, Device::CPU>& matrix) {
      +
      244  const Distribution& dist = matrix.distribution();
      +
      245  for (auto tile_wrt_local : iterate_range2d(dist.localNrTiles())) {
      +
      246  GlobalTileIndex tile_wrt_global = dist.globalTileIndex(tile_wrt_local);
      +
      247  auto tl_index = dist.globalElementIndex(tile_wrt_global, {0, 0});
      +
      248  auto seed = tl_index.col() + tl_index.row() * matrix.size().cols();
      +
      249 
      +
      250  using TileType = typename std::decay_t<decltype(matrix)>::TileType;
      +
      251  auto rnd_f = [seed](TileType&& tile) {
      +
      252  internal::getter_random<T> random_value(seed);
      +
      253  for (auto el_idx : iterate_range2d(tile.size())) {
      +
      254  tile(el_idx) = random_value();
      +
      255  }
      +
      256  };
      +
      257 
      +
      258  dlaf::internal::transformDetach(dlaf::internal::Policy<Backend::MC>(), std::move(rnd_f),
      +
      259  matrix.readwrite(tile_wrt_local));
      +
      260  }
      +
      261 }
      +
      262 
      +
      263 namespace internal {
      +
      264 
      +
      265 template <class T>
      +
      266 void set_diagonal_tile(const Tile<T, Device::CPU>& tile, internal::getter_random<T>& random_value,
      +
      267  SizeType offset_value) {
      +
      268  // DIAGONAL
      +
      269  // for diagonal tiles get just lower matrix values and set value for both
      +
      270  // straight and transposed indices
      +
      271  for (SizeType j = 0; j < tile.size().cols(); ++j) {
      +
      272  for (SizeType i = 0; i < j; ++i) {
      +
      273  auto value = random_value();
      +
      274 
      +
      275  tile(TileElementIndex{i, j}) = value;
      +
      276  tile(TileElementIndex{j, i}) = dlaf::conj(value);
      +
      277  }
      +
      278  tile(TileElementIndex{j, j}) = std::real(random_value()) + offset_value;
      +
      279  }
      +
      280 }
      +
      281 
      +
      282 template <class T>
      +
      283 void set_lower_and_upper_tile(const Tile<T, Device::CPU>& tile, internal::getter_random<T>& random_value,
      +
      284  TileElementSize full_tile_size, GlobalTileIndex tile_wrt_global) {
      +
      285  // LOWER or UPPER (except DIAGONAL)
      +
      286  // random values are requested in the same order for both original and transposed
      +
      287  for (SizeType j = 0; j < full_tile_size.cols(); ++j) {
      +
      288  for (SizeType i = 0; i < full_tile_size.rows(); ++i) {
      +
      289  auto value = random_value();
      +
      290 
      +
      291  // but they are set row-wise in the original tile and col-wise in the
      +
      292  // transposed one
      +
      293  if (tile_wrt_global.row() > tile_wrt_global.col()) { // LOWER
      +
      294  TileElementIndex index{i, j};
      +
      295  if (index.isIn(tile.size()))
      +
      296  tile(index) = value;
      +
      297  }
      +
      298  else { // UPPER
      +
      299  TileElementIndex index{j, i};
      +
      300  if (index.isIn(tile.size()))
      +
      301  tile(index) = dlaf::conj(value);
      +
      302  }
      +
      303  }
      +
      304  }
      +
      305 }
      +
      306 
      +
      323 template <class T>
      +
      324 void set_random_hermitian_with_offset(Matrix<T, Device::CPU>& matrix, const SizeType offset_value) {
      +
      325  // note:
      +
      326  // By assuming square blocksizes, it is easier to locate elements. In fact:
      +
      327  // - Elements on the diagonal are stored in the diagonal of the diagonal tiles
      +
      328  // - Tiles under the diagonal store elements of the lower triangular matrix
      +
      329  // - Tiles over the diagonal store elements of the upper triangular matrix
      +
      330 
      +
      331  const Distribution& dist = matrix.distribution();
      +
      332 
      +
      333  DLAF_ASSERT(square_size(matrix), matrix);
      +
      334  DLAF_ASSERT(square_blocksize(matrix), matrix);
      +
      335 
      +
      336  auto full_tile_size = matrix.blockSize();
      +
      337 
      +
      338  for (auto tile_wrt_local : iterate_range2d(dist.localNrTiles())) {
      +
      339  GlobalTileIndex tile_wrt_global = dist.globalTileIndex(tile_wrt_local);
      +
      340 
      +
      341  auto tl_index = dist.globalElementIndex(tile_wrt_global, {0, 0});
      +
      342 
      +
      343  // compute the same seed for original and "transposed" tiles, so transposed ones will know the
      +
      344  // values of the original one without the need of accessing real values (nor communication in case
      +
      345  // of distributed matrices)
      +
      346  SizeType seed;
      +
      347  if (tile_wrt_global.row() >= tile_wrt_global.col()) // LOWER or DIAGONAL
      +
      348  seed = tl_index.col() + tl_index.row() * matrix.size().cols();
      +
      349  else
      +
      350  seed = tl_index.row() + tl_index.col() * matrix.size().rows();
      +
      351 
      +
      352  using TileType = typename std::decay_t<decltype(matrix)>::TileType;
      +
      353  auto set_hp_f = [=](const TileType& tile) {
      +
      354  internal::getter_random<T> random_value(seed);
      +
      355  if (tile_wrt_global.row() == tile_wrt_global.col())
      +
      356  internal::set_diagonal_tile(tile, random_value, offset_value);
      +
      357  else
      +
      358  internal::set_lower_and_upper_tile(tile, random_value, full_tile_size, tile_wrt_global);
      +
      359  };
      +
      360 
      +
      361  dlaf::internal::transformDetach(dlaf::internal::Policy<Backend::MC>(), std::move(set_hp_f),
      +
      362  matrix.readwrite(tile_wrt_local));
      +
      363  }
      +
      364 }
      +
      365 
      +
      366 }
      +
      367 
      +
      381 template <class T>
      +
      382 void set_random_hermitian(Matrix<T, Device::CPU>& matrix) {
      +
      383  internal::set_random_hermitian_with_offset(matrix, 0);
      +
      384 }
      +
      385 
      +
      402 template <class T>
      +
      403 void set_random_hermitian_positive_definite(Matrix<T, Device::CPU>& matrix) {
      +
      404  internal::set_random_hermitian_with_offset(matrix, 2 * matrix.size().rows());
      +
      405 }
      +
      406 
      +
      407 }
      +
      408 }
      +
      409 }
      enum_output.h
      dlaf::comm::CommunicatorGrid
      Definition: communicator_grid.h:42
      @@ -392,11 +397,11 @@
      dlaf::matrix::Distribution::globalElementIndex
      GlobalElementIndex globalElementIndex(const GlobalTileIndex &global_tile, const TileElementIndex &tile_element) const noexcept
      Definition: distribution.h:176
      dlaf::matrix::Distribution::localNrTiles
      const LocalTileSize & localNrTiles() const noexcept
      Returns the number of tiles stored locally (2D size).
      Definition: distribution.h:147
      dlaf::matrix::Distribution::globalTileIndex
      GlobalTileIndex globalTileIndex(const GlobalElementIndex &global_element) const noexcept
      Definition: distribution.h:189
      -
      dlaf::matrix::Matrix< const T, D >
      Definition: matrix.h:151
      -
      dlaf::matrix::Matrix
      Definition: matrix.h:55
      -
      dlaf::matrix::Matrix::readwrite
      ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
      Definition: matrix.h:121
      +
      dlaf::matrix::Matrix< const T, D >
      Definition: matrix.h:171
      +
      dlaf::matrix::Matrix
      Definition: matrix.h:56
      +
      dlaf::matrix::Matrix::readwrite
      ReadWriteSenderType readwrite(const LocalTileIndex &index) noexcept
      Definition: matrix.h:122
      dlaf::matrix::Tile
      Definition: tile.h:329
      -
      dlaf::matrix::util::internal::getter_random
      Callable that returns random values in the range [-1, 1].
      Definition: util_matrix.h:107
      +
      dlaf::matrix::util::internal::getter_random
      Callable that returns random values in the range [-1, 1].
      Definition: util_matrix.h:113
      @@ -406,20 +411,21 @@
      dlaf::matrix::Panel::readwrite
      ReadWriteSenderType readwrite(LocalTileIndex index)
      Definition: panel.h:570
      dlaf::conj
      std::complex< T > conj(const std::complex< T > number)
      Return complex conjugate of a complex number.
      Definition: types.h:165
      -
      dlaf::matrix::equal_blocksize
      bool equal_blocksize(const Matrix< const T, D1 > &lhs, Matrix< const T, D2 > &rhs) noexcept
      Returns true if matrices have equal blocksizes.
      Definition: util_matrix.h:60
      -
      dlaf::matrix::multipliable
      bool multipliable(const Matrix< const T, D > &a, const Matrix< const T, D > &b, const Matrix< const T, D > &c, const blas::Op opA, const blas::Op opB) noexcept
      Returns true if matrices a, b and c have matrix multipliable sizes and block sizes.
      Definition: util_matrix.h:96
      -
      dlaf::matrix::util::set
      void set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f)
      Definition: util_matrix.h:178
      -
      dlaf::matrix::util::internal::set_random_hermitian_with_offset
      void set_random_hermitian_with_offset(Matrix< T, Device::CPU > &matrix, const SizeType offset_value)
      Definition: util_matrix.h:318
      +
      dlaf::matrix::equal_blocksize
      bool equal_blocksize(const Matrix< const T, D1 > &lhs, Matrix< const T, D2 > &rhs) noexcept
      Returns true if matrices have equal blocksizes.
      Definition: util_matrix.h:66
      +
      dlaf::matrix::multipliable
      bool multipliable(const Matrix< const T, D > &a, const Matrix< const T, D > &b, const Matrix< const T, D > &c, const blas::Op opA, const blas::Op opB) noexcept
      Returns true if matrices a, b and c have matrix multipliable sizes and block sizes.
      Definition: util_matrix.h:102
      +
      dlaf::matrix::util::set
      void set(Matrix< T, Device::CPU > &matrix, ElementGetter el_f)
      Definition: util_matrix.h:184
      +
      dlaf::matrix::util::internal::set_random_hermitian_with_offset
      void set_random_hermitian_with_offset(Matrix< T, Device::CPU > &matrix, const SizeType offset_value)
      Definition: util_matrix.h:324
      dlaf::matrix::square_blocksize
      bool square_blocksize(const MatrixLike &m) noexcept
      Returns true if the matrix block size is square.
      Definition: util_matrix.h:48
      -
      dlaf::matrix::util::set_random_hermitian
      void set_random_hermitian(Matrix< T, Device::CPU > &matrix)
      Definition: util_matrix.h:376
      -
      dlaf::matrix::equal_process_grid
      bool equal_process_grid(const Matrix< const T, D > &m, const comm::CommunicatorGrid &g) noexcept
      Returns true if the matrix is distributed on the communication grid.
      Definition: util_matrix.h:72
      -
      dlaf::matrix::local_matrix
      bool local_matrix(const Matrix< const T, D > &m) noexcept
      Returns true if the matrix is local to a process.
      Definition: util_matrix.h:66
      -
      dlaf::matrix::equal_distributions
      bool equal_distributions(const Matrix< const T, D1 > &lhs, const Matrix< const T, D2 > &rhs) noexcept
      Returns true if the matrices are distributed the same way.
      Definition: util_matrix.h:78
      -
      dlaf::matrix::util::set_random
      void set_random(Matrix< T, Device::CPU > &matrix)
      Definition: util_matrix.h:237
      -
      dlaf::matrix::multipliable_sizes
      bool multipliable_sizes(common::Size2D< IndexT, Tag > a, common::Size2D< IndexT, Tag > b, common::Size2D< IndexT, Tag > c, const blas::Op opA, const blas::Op opB) noexcept
      Returns true if the sizes are compatible for matrix multiplication.
      Definition: util_matrix.h:84
      +
      dlaf::matrix::util::set_random_hermitian
      void set_random_hermitian(Matrix< T, Device::CPU > &matrix)
      Definition: util_matrix.h:382
      +
      dlaf::matrix::single_tile_per_block
      bool single_tile_per_block(const MatrixLike &m) noexcept
      Returns true if the matrix has a single tile per block.
      Definition: util_matrix.h:54
      +
      dlaf::matrix::equal_process_grid
      bool equal_process_grid(const Matrix< const T, D > &m, const comm::CommunicatorGrid &g) noexcept
      Returns true if the matrix is distributed on the communication grid.
      Definition: util_matrix.h:78
      +
      dlaf::matrix::local_matrix
      bool local_matrix(const Matrix< const T, D > &m) noexcept
      Returns true if the matrix is local to a process.
      Definition: util_matrix.h:72
      +
      dlaf::matrix::equal_distributions
      bool equal_distributions(const Matrix< const T, D1 > &lhs, const Matrix< const T, D2 > &rhs) noexcept
      Returns true if the matrices are distributed the same way.
      Definition: util_matrix.h:84
      +
      dlaf::matrix::util::set_random
      void set_random(Matrix< T, Device::CPU > &matrix)
      Definition: util_matrix.h:243
      +
      dlaf::matrix::multipliable_sizes
      bool multipliable_sizes(common::Size2D< IndexT, Tag > a, common::Size2D< IndexT, Tag > b, common::Size2D< IndexT, Tag > c, const blas::Op opA, const blas::Op opB) noexcept
      Returns true if the sizes are compatible for matrix multiplication.
      Definition: util_matrix.h:90
      dlaf::matrix::square_size
      bool square_size(const MatrixLike &m) noexcept
      Returns true if the matrix is square.
      Definition: util_matrix.h:42
      -
      dlaf::matrix::equal_size
      bool equal_size(const MatrixLikeA &lhs, const MatrixLikeB &rhs) noexcept
      Returns true if matrices have equal sizes.
      Definition: util_matrix.h:54
      -
      dlaf::matrix::util::set_random_hermitian_positive_definite
      void set_random_hermitian_positive_definite(Matrix< T, Device::CPU > &matrix)
      Definition: util_matrix.h:397
      +
      dlaf::matrix::equal_size
      bool equal_size(const MatrixLikeA &lhs, const MatrixLikeB &rhs) noexcept
      Returns true if matrices have equal sizes.
      Definition: util_matrix.h:60
      +
      dlaf::matrix::util::set_random_hermitian_positive_definite
      void set_random_hermitian_positive_definite(Matrix< T, Device::CPU > &matrix)
      Definition: util_matrix.h:403